University of Wollongong Thesis Collections
University of Wollongong Thesis Collection
University of Wollongong Year
Boundary element formulations for
fracture mechanics problems
Wei-Liang WuUniversity of Wollongong
Wu, Wei-Liang, Boundary element formulations for fracture mechanics problems, PhDthesis, School of Mathematics and Applied Statistics, University of Wollongong, 2004.http://ro.uow.edu.au/theses/253
This paper is posted at Research Online.
http://ro.uow.edu.au/theses/253
BOUNDARY ELEMENT FORMULATIONS FOR
FRACTURE MECHANICS PROBLEMS
By
Wei-Liang Wu
SUBMITTED IN PARTIAL FULFILLMENT OF THE
REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
AT
UNIVERSITY OF WOLLONGONG
NORTHFIELDS AVENUE, WOLLONGONG NSW 2522
AUSTRALIA
NOVEMBER 2004
UNIVERSITY OF WOLLONGONG
SCHOOL OF
MATHEMATICS AND APPLIED STATICS
This thesis is submitted to the University of Wollongong, and
has not been submitted for a higher degree to any other university or
institution.
Wei-Liang Wu
November 2004
ii
UNIVERSITY OF WOLLONGONG
Date: November 2004
Author: Wei-Liang Wu
Title: Boundary Element Formulations for Fracture
Mechanics Problems
School: Mathematics and Applied Statics
Degree: Ph.D.
Permission is herewith granted to University of Wollongong to circulate
and to have copied for non-commercial purposes, at its discretion, the above
title upon the request of individuals or institutions.
Signature of Author
THE AUTHOR RESERVES OTHER PUBLICATION RIGHTS, ANDNEITHER THE THESIS NOR EXTENSIVE EXTRACTS FROM IT MAYBE PRINTED OR OTHERWISE REPRODUCED WITHOUT THE AUTHOR’SWRITTEN PERMISSION.
THE AUTHOR ATTESTS THAT PERMISSION HAS BEEN OBTAINEDFOR THE USE OF ANY COPYRIGHTED MATERIAL APPEARING IN THISTHESIS (OTHER THAN BRIEF EXCERPTS REQUIRING ONLY PROPERACKNOWLEDGEMENT IN SCHOLARLY WRITING) AND THAT ALL SUCH USEIS CLEARLY ACKNOWLEDGED.
iii
To My Parents with Love and Gratitude.
iv
Acknowledgements
I would like to acknowledge the many people who have helped me in my life and who
helped bring this thesis to fruition.
In particular, I want to thank my supervisor, Dr. Xiaoping Lu who had constantly
been a source of knowledge and expertise on mathematical techniques which have been
utilised in this thesis. I thank her for her constant support, encouragement and her
willingness to help at any time despite her hectic schedule.
Above all, I want to thank my teachers – both in Taiwan and Australia. To those
who helped me learn in the classroom. To those who helped me learn in the world
beyond the classroom.
I would also like to thank the colleagues of the University of Wollongong who
devoted time and effort to provide me with critical feedback on earlier drafts, point-
ing inaccuracies, suggesting better examples, and gently noting things that needed
smoothing. I would like to mention especially Dr. Ahmed A. El-Feki, Dr. Anna Maria
Milan, Dr. Bin Liu and Frank Bierbrauer.
In addition, my deepest thanks are owed to my family who have always inspired
me to greater things particularly my parents for their love, sacrifice and support.
v
Table of Contents
Acknowledgements v
List of Tables ix
List of Figures x
1 Introduction 1
1.1 General Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Modes of Crack Tip Deformation . . . . . . . . . . . . . . . . . . . . 4
1.3 Other Fracture Characterising Parameters . . . . . . . . . . . . . . . 6
1.3.1 The J-integral . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.2 The Crack Tip Opening Displacement . . . . . . . . . . . . . 7
1.4 Numerical Methods in Linear Elastic Fracture Mechanics . . . . . . . 8
1.4.1 Finite Difference Method . . . . . . . . . . . . . . . . . . . . . 8
1.4.2 Finite Element Method . . . . . . . . . . . . . . . . . . . . . . 8
1.4.3 Boundary Element Method . . . . . . . . . . . . . . . . . . . 9
1.5 Advanced Formulations in Boundary Element Method . . . . . . . . . 11
1.5.1 Dual Boundary Element Method . . . . . . . . . . . . . . . . 11
1.5.2 Subregion Boundary Element Method . . . . . . . . . . . . . . 12
1.5.3 Dual Reciprocity Boundary Element Method . . . . . . . . . . 13
2 Dual Boundary Integral Formulation for Two Dimensional Crack
Problems 15
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2 The Dual Boundary Integral Formulation . . . . . . . . . . . . . . . . 18
2.3 Stress Intensity Factor Calculation . . . . . . . . . . . . . . . . . . . 22
2.4 Numerical Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
vi
3 An Efficient Dual Boundary Element Method for Crack Problems
with Anti-plane Shear Loading 43
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.2 The Dual Boundary Integral Equation
for Anti-plane Problems . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.3 Calculation of the Mode III Stress Intensity Factor . . . . . . . . . . 49
3.4 Numerical Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4 The Evaluation of Stress Intensity Factors 60
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2 The Numerical Evaluation of Stress Intensity Factors . . . . . . . . . 63
4.2.1 J-integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.2.2 The Discontinuous Quarter Point Element Method . . . . . . 66
4.2.3 The Special Crack Tip Element Method . . . . . . . . . . . . 69
4.3 Numerical Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.3.1 A Plate with a Central Slant Crack . . . . . . . . . . . . . . . 71
4.3.2 Infinite Plate with Two Inclined Cracks . . . . . . . . . . . . . 77
4.3.3 Infinite Plate with Two Parallel Cracks . . . . . . . . . . . . . 82
4.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5 A New Subregion Boundary Element Technique 87
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.2 The Multi Region Technique of Boundary Element Method . . . . . . 91
5.3 Comparison of Subregion BEM Techniques . . . . . . . . . . . . . . . 94
5.3.1 The Traditional Method . . . . . . . . . . . . . . . . . . . . . 94
5.3.2 Kita & Kamiya’s Method . . . . . . . . . . . . . . . . . . . . 94
5.3.3 The Proposed Method . . . . . . . . . . . . . . . . . . . . . . 96
5.4 Numerical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6 A Subregion DRBEM Formulation for the Dynamic Analysis of Two
Dimensional Cracks 109
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
6.2 Dual Reciprocity Boundary Element Method . . . . . . . . . . . . . . 113
6.3 Derivation of Particular Solutions . . . . . . . . . . . . . . . . . . . . 118
6.4 The Dynamic Stress Intensity Factors . . . . . . . . . . . . . . . . . . 119
6.5 Numerical Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
6.5.1 A Rectangular Plate with a Central Crack . . . . . . . . . . . 121
vii
6.5.2 A Rectangular Plate with a Central Slant Crack . . . . . . . . 123
6.5.3 A Rectangular Plate with an Internal Kinked Crack . . . . . . 126
6.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
7 Conclusion 132
Bibliography 135
viii
List of Tables
2.1 Mode I SIF for an internal kinked crack in a rectangular plate . . . . 29
2.2 Mode II SIF for an internal kinked crack in a rectangular plate . . . . 29
2.3 Comparison of normalised stress intensity factors . . . . . . . . . . . 35
3.1 Normalised mode III stress intensity factor for a straight central crack 52
ix
List of Figures
1.1 The three modes of loading . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Definition of the coordinate axis ahead of a crack tip . . . . . . . . . 5
1.3 Stress normal to the crack plane in mode I . . . . . . . . . . . . . . . 6
2.1 Rectangular plate with a central slant crack (h/w = 2, θ = 45) . . . . 23
2.2 Normalised mode I SIF for the rectangular plate with a central slant
crack: (a) the present method, (b) Reference [80], and (c) Reference [73] 24
2.3 Normalised mode II SIF for the rectangular plate with a central slant
crack: (a) the present method, (b) Reference [80], and (c) Reference [73] 25
2.4 Normalised mode I SIF for the rectangular plate with a central slant
crack (a/w = 0.1): (a) the present method, and (b) Reference [73] . . 26
2.5 Normalised mode II SIF for the rectangular plate with a central slant
crack (a/w = 0.1): (a) the present method, and (b) Reference [73] . . 27
2.6 Rectangular plate with an internal kinked crack (h/w = 2, a/w = 0.1) 28
2.7 Normalised mode I SIF vs. crack ratio b/a at tip A. The angle of the
kinked crack is (a) 30, (b) 45, (c) 60 . . . . . . . . . . . . . . . . . 30
2.8 Normalised mode II SIF vs. crack ratio b/a at tip A. The angle of the
kinked crack is (a) 30, (b) 45, (c) 60 . . . . . . . . . . . . . . . . . 31
2.9 Normalised mode I SIF vs. crack ratio b/a at tip B. The angle of the
kinked crack is (a) 30, (b) 45, (c) 60 . . . . . . . . . . . . . . . . . 32
2.10 Normalised mode II SIF vs. crack ratio b/a at tip B. The angle of the
kinked crack is (a) 30, (b) 45, (c) 60 . . . . . . . . . . . . . . . . . 33
2.11 Rectangular plate with two inclined cracks (h/w = 2, a/W = 0.25) . . 34
2.12 Normalised mode I SIF vs. inclined angle θ: (a) Finite region (b) In-
finite region at tip A, and (c) Finite region (d) Infinite region at tip
B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
x
2.13 Normalised mode II SIF vs. inclined angle θ: (a) Finite region (b)
Infinite region at tip A, and (c) Finite region (d) Infinite region at tip B 37
2.14 Rectangular plate with two parallel cracks (h/w = 2, a/w = 0.025) . . 38
2.15 Normalised mode I SIF for the rectangular plate with two parallel
cracks: (a) the present method, and (b) Reference [93] . . . . . . . . 39
2.16 Normalised mode I SIF vs. s, with a/w given by (a) 0.025, (b) 0.05,
(c) 0.1, (d) 0.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.17 Normalised mode II SIF vs. s, with a/w given by (a) 0.025, (b) 0.05,
(c) 0.1, (d) 0.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.1 Rectangular plate with a central slant crack . . . . . . . . . . . . . . 51
3.2 Normalised mode III stress intensity factor (SIF) for the rectangular
plate with a central slant crack (a) θ = 30, (b) θ = 45, and (c) θ = 60 53
3.3 Normalised mode III SIF for the infinite plate with a central slant crack
(a) the analytical solutions, and (b) the present method . . . . . . . 54
3.4 A finite plate with two collinear cracks . . . . . . . . . . . . . . . . . 55
3.5 Normalised mode III SIF for the rectangular plate with two identical
collinear cracks at tip A: (a) the analytical results, (b) the present
method and at tip B: (c) the analytical solutions, (d) the present
method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.6 A finite plate with two parallel cracks . . . . . . . . . . . . . . . . . 57
3.7 Normalised mode III SIF for the rectangular plate with two parallel
cracks (a) Reference [93] (b) the present method . . . . . . . . . . . 58
4.1 Crack in an infinite plane . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2 General continuous quadratic element . . . . . . . . . . . . . . . . . . 63
4.3 Coordinate reference system and contour path for J-integral . . . . . 64
4.4 Singular quarter-point boundary elements . . . . . . . . . . . . . . . 66
4.5 Modeling of the quarter point boundary element . . . . . . . . . . . . 68
4.6 Relative error of Mode I SIF for the infinite plate with a central slant
crack from the QPE with (a) 6 elements, (b) 8 elements, (c) 10 elements
and the SCT with (d) 6 elements, (e) 8 elements, and (f) 10 elements 72
4.7 Relative error of Mode II SIF for the infinite plate with a central slant
crack from the QPE with (a) 6 elements, (b) 8 elements, (c) 10 elements
and the SCT with (d) 6 elements, (e) 8 elements, and (f) 10 elements 73
xi
4.8 Relative error of Mode III SIF for the infinite plate with a central
slant crack from the QPE with (a) 6 elements, (b) 8 elements, (c) 10
elements and the SCT with (d) 6 elements, (e) 8 elements, and (f) 10
elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.9 Mode I SIF of the finite plate with a central slant crack from the QPE
with aw
= (a) 23, (b) 1
2, (c) 1
5and the SCT with (d) 2
3, (e) 1
2, (f) 1
5. . . 75
4.10 Mode II SIF of the finite plate with a central slant crack from the
QPE with aw
= (a) 23, (b) 1
2, (c) 1
5and the SCT with (d) 2
3, (e) 1
2, (f) 1
576
4.11 Relative error of Mode I SIF for the infinite plate with two inclined
cracks at tip A from the QPE with (a) 6 elements, (b) 8 elements, (c)
10 elements and the SCT with (d) 6 elements, (e) 8 elements, and (f)
10 elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.12 Relative error of Mode I SIF for the infinite plate with two inclined
cracks at tip B from the QPE with (a) 6 elements, (b) 8 elements, (c)
10 elements and the SCT with (d) 6 elements, (e) 8 elements, and (f)
10 elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.13 Relative error of Mode III SIF for the infinite plate with two inclined
cracks at tip A from the QPE with (a) 6 elements, (b) 8 elements, (c)
10 elements and the SCT with (d) 6 elements, (e) 8 elements, and (f)
10 elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.14 Relative error of Mode III SIF for the infinite plate with two inclined
cracks at tip B from the QPE with (a) 6 elements, (b) 8 elements, (c)
10 elements and the SCT with (d) 6 elements, (e) 8 elements, and (f)
10 elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.15 Mode I SIF for the infinite plate with two parallel cracks from (a) the
Reference [93], the QPE with (b) 6 elements, (c) 8 elements, (d) 10
elements and the SCT with (e) 6 elements, (f) 8 elements, and (g) 10
elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.16 Mode II SIF for the infinite plate with two parallel cracks from the
QPE with (a) 6 elements, (b) 8 elements, (c) 10 elements and the SCT
with (d) 6 elements, (e) 8 elements, and (f) 10 elements . . . . . . . 84
4.17 Mode III SIF for the infinite plate with two parallel cracks from (a)
the Reference [93], the QPE with (b) 6 elements, (c) 8 elements, (d)
10 elements and the SCT with (e) 6 elements, (f) 8 elements, and (g)
10 elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
xii
5.1 A three subregion medium . . . . . . . . . . . . . . . . . . . . . . . . 92
5.2 A perfectly bonded dissimilar elastic semi-strip . . . . . . . . . . . . . 101
5.3 Normalised normal stress distribution on the interface
(E2/E1, ν1, ν2) = (9.0, 0.5, 0.5): (a) Ref. [17], (b) the present method
(E2/E1, ν1, ν2) = (3.0, 0.5, 0.5): (c) Ref. [17], (d) the present method . 102
5.4 A three-layer plate with a centre crack . . . . . . . . . . . . . . . . . 103
5.5 Normalised mode I stress intensity factor (SIF) on the three-layer plate
with a centre crack: (a) l/h2 = 0.1, (b) l/h2 = 0.2, (c) l/h2 = 0.3, (d)
l/h2 = 0.4, (e) l/h2 = 0.5 . . . . . . . . . . . . . . . . . . . . . . . . 104
5.6 A three-layer plate with two identical co-linear cracks . . . . . . . . . 105
5.7 Normalised mode I SIF at tip A on the three-layer plate with two
co-linear cracks: (a) l/h2 = 0.1, (b) l/h2 = 0.2, (c) l/h2 = 0.3, (d)
l/h2 = 0.4, (e) l/h2 = 0.5 . . . . . . . . . . . . . . . . . . . . . . . . 106
5.8 Normalised mode I SIF at tip B on the three-layer plate with two
co-linear cracks: (a) l/h2 = 0.1, (b) l/h2 = 0.2, (c) l/h2 = 0.3, (d)
l/h2 = 0.4, (e) l/h2 = 0.5 . . . . . . . . . . . . . . . . . . . . . . . . 107
6.1 Boundary and internal nodes . . . . . . . . . . . . . . . . . . . . . . . 114
6.2 Crack on the body for subregion method . . . . . . . . . . . . . . . . 118
6.3 Rectangular plate with a central crack . . . . . . . . . . . . . . . . . 121
6.4 Normalised mode I DSIF for the rectangular plate with a central crack
(a) the present method, (b) the dual reciprocity method [6] and (c) the
Laplace transform method [6] . . . . . . . . . . . . . . . . . . . . . . 122
6.5 Rectangular plate with a central slant crack . . . . . . . . . . . . . . 123
6.6 Normalised mode I DSIF for the rectangular plate with a central slant
crack (a) the present method, (b) the dual reciprocity method [6] and
(c) the Laplace transform method [6] . . . . . . . . . . . . . . . . . . 124
6.7 Normalised mode II DSIF for the rectangular plate with a central slant
crack (a) the present method, (b) the dual reciprocity method [6] and
(c) the Laplace transform method [6] . . . . . . . . . . . . . . . . . . 125
6.8 Rectangular plate with an internal kinked crack . . . . . . . . . . . . 126
6.9 Normalised mode I DSIF for the rectangular plate with an internal
kinked crack at tip A (a) b/a = 0.2, (b) b/a = 0.4, and (c) b/a = 0.6 . 128
6.10 Normalised mode I DSIF for the rectangular plate with an internal
kinked crack at tip B (a) b/a = 0.2, (b) b/a = 0.4, and (c) b/a = 0.6 . 129
xiii
6.11 Normalised mode II DSIF for the rectangular plate with an internal
kinked crack at tip B (a) b/a = 0.2, (b) b/a = 0.4, and (c) b/a = 0.6 . 130
xiv
Chapter 1
Introduction
1.1 General Considerations
The design of modern structures necessitates the consideration of fractures. Iron and
steel saw increasing structural use in the 19th century and fracture was a problem.
In Great Britain, there were two hundred people killed per year in railway accidents
during the period 1860-1870 [22]. Most of the accidents were caused by fractures of
wheels, axles or rails, and were detrimental to the economy. In 1861, a leading article
of “The Engineer” reported: “Effects of percussion and frost upon iron... . We need
hardly say that this is one of the most important subjects that engineers of the present
day are called upon to investigate. The lives of many persons, and the property of
many more, will be saved if the truth of the matter be discovered – lost if it be not”
[15]. More time and effort needs to be spent on maintenance and repair of structures
in order to minimize the potential of component fracture and subsequent structure
failure. An economic study estimated the cost of fracture in the United States in
1978 at approximately US$119 billion, about 4% of the gross national product [35].
Some of these failures occurred due to poor designs, but many were due to material
1
Chapter 1: Introduction 2
deficiencies in the form of pre-existing flaws that initiated cracks and thus caused
fractures.
However, no material is flawless and fabrication flaws cannot be avoided. From
a practical point of view, many cracks may be considered as harmless. A structure,
such as a tanker, contains probably several thousand macroscopic cracks and several
million micro-cracks [21]. Essentially, if the cracks reach a certain size and situated in
a highly strained region, they are regarded as a potential cause of fractures. There-
fore, engineers must understand and characterise cracks and their effects, and try to
predict if and when they may become unsafe during the structures operational service
life. The subject of this thesis is to examine numerical techniques that can be used
to efficiently solve fracture problems. These techniques can then aid engineers in
studying and understanding the characteristics of cracks under particular conditions.
Griffith [49], in 1920, was the first systematic investigator of fracture by experi-
mental studies. In order to explain the measured strengths of the crack propagation,
Griffith introduced a model where the global energy balance controls the growth of
cracks. Fracture initiates when the energy stored in the fracture overcomes the surface
energy of the material. This theory is valid for materials with little or no preceding
plastic deformation near the crack tip and the fracture strength depends on flaw size.
From 1940, a major development of fracture theory was the brittle fracture of large
welded structures such as bridges, ships and oil storage containers. Linear elastic
fracture mechanics (LEFM) succeeded in application to the reliability assessment of
aircrafts. A fatigue crack that grew from an astro-navigation window caused the loss
of two aircrafts in 1953 and 1954. LEFM could provide an explanation for the fracture
[55, 96].
Chapter 1: Introduction 3
In 1956, Irwin [56] developed a new formulation of energy release rate and its
critical value by extending Griffith’s theory for ideally brittle materials. The energy
release rate is defined as the rate of change in potential energy with crack area for a
linear elastic material, and its critical value is a measure of fracture toughness. Thus,
Irwin postulated that crack growth occurs when the energy release rate reaches its
critical value. Another important contribution by Irwin [57], in 1957, was that the
distribution of the crack tip stress field was recognized and solved by the mathematical
procedures of Westergaard [101]. It is shown that the stresses and displacements near
the crack tip could be described by a single constant that was related to the energy
release rate. This crack tip characterising parameter is known as the stress intensity
factor. The stress intensity factor is a measure of the intensity of the singularity in
the stress field near the crack tip. It plays a fundamental role in linear elastic fracture
mechanics anaysis.
In 1988, a Boeing 737 lost a top section of the fuselage over Hawaii and yet it still
managed to land safely. The cause of this fatigue failure was multi-site fatigue, which
is still poorly understood. In 2001, American Airlines Flight 587, an airbus A300,
crashed in Queens after taking off from Kennedy International Airport, New York.
According to a National Transportation Safety Board statement, the plane’s vertical
stabilizer and rudder separated from the fuselage before impact. It has become the
subject of investigator’s scrutiny due to the pre-crash separation. But the cause of
the flight crash remains unknown. There is no doubt that there will be more and
more applications of fracture mechanics.
Chapter 1: Introduction 4
1.2 Modes of Crack Tip Deformation
The fundamental postulate of linear elastic fracture mechanics is based on the theory
that the intensity of the stress or strain fields surrounding a crack tip may be uniquely
described in terms of the stress intensity factor. A crack can be loaded in three
different modes, see Figure 1.1. These modes represent the local deformation ahead
Figure 1.1: The three modes of loading
of a crack. Normal stresses give rise to the opening mode or mode I loading. The
displacements of the crack surfaces are perpendicular to the plane of the crack. In-
plane shear results in mode II or sliding mode. The displacement of the crack surfaces
slide over each other along the crack line. The tearing or anti-plane mode, or mode
III, is caused by out-of-plane shear. The displacement of the crack surfaces slide over
each other perpendicular to the crack line. The superposition of these three modes is
sufficient to describe the most general case of crack surface displacement.
Irwin showed that the stress field in any linear elastic cracked body is given by
Chapter 1: Introduction 5
the following series representation of the stress tensor σij:
σij =1√r
KIfIij(θ) + KIIf
IIij (θ) + KIIIf
IIIij (θ) + higher order terms
(1.1)
where r and θ are as defined in Figure 1.2, KI , KII and KIII denote stress intensity
Figure 1.2: Definition of the coordinate axis ahead of a crack tip
factors corresponding to the three basic modes of crack surface displacement. The
higher order terms depend on geometry, but the solution for any given configuration
contains a leading term that is proportional to 1/√
r. As r → 0, the leading term
approaches infinity, but the other terms remain finite or approach zero. Thus stress
near the crack tip varies with 1/√
r, regardless of the configuration of the cracked
body. It can also be shown that displacement near the crack tip varies with√
r. In
equation (1.1), it is seen that r → 0 leads to a stress singularity, as the stress tensor
σij tends to infinity. This is shown in Figure 1.3.
Chapter 1: Introduction 6
Figure 1.3: Stress normal to the crack plane in mode I
1.3 Other Fracture Characterising Parameters
Two other fracture mechanics parameters that can be used to calculate the stress
intensity factor are the J-integral, and the crack tip opening displacement.
1.3.1 The J-integral
The J-integral represents the energy extracted through the crack tip singularity. The
original concept of the path independent J-integral was developed by Rice [81].
The path independent J−integral characterises the stress-strain field at the crack
tip, whose path is taken sufficiently far from the crack tip for the cracks to be analysed
elastically, where singularities or non-linear elasto-plastic behaviour are not encoun-
tered.
In this method, the physical interpretation in terms of the potential energy is
Chapter 1: Introduction 7
available for crack extension, but retains its physical significance as a measure of
the intensity of the characteristic crack tip strain field. This is similar to the stress
intensity factor concept, which measures the intensity of the stress field in the vicinity
of the crack tip.
1.3.2 The Crack Tip Opening Displacement
In 1963, Wells [97] proposed that the fracture behaviour in the vicinity of a sharp
crack could be characterised by the crack tip opening displacement. Furthermore,
Wells showed that the concept of crack tip opening displacement was analogous to
the concept of critical crack extension force. It is postulated that failure occurs when
the crack tip opening displacement value reaches a critical value on the assumption
that this value characterises the strain or stress ahead of the crack tip necessary to
initiate the failure mechanisms. Accordingly, it is the plastic strain in the vicinity
of the crack tip region that controls fracture. A measure of the amount of crack
tip plastic strain is the displacement of the crack tip, especially at or very close
to the tip. Thus, it might be expected that, at the onset of fracture, crack tip
opening displacement has a critical value for a given material and accordingly can
be used as a fracture characterising parameter. Burdekin and Stone [23], in 1966,
used Dugdale’s strip yield model [36] and provided an improved basis for the crack
tip opening displacement concept. Burdekin and Stone have also shown that under
linear elastic fracture mechanics conditions the crack tip opening displacement could
be related to the stress intensity factors. Therefore, this technique is compatible with
the stress intensity factor approach.
Chapter 1: Introduction 8
1.4 Numerical Methods in Linear Elastic Fracture
Mechanics
The solution of engineering problems can be obtained using analytical methods or
numerical methods. Analytical solutions of fracture problems are limited to idealised
situations wherein the domain is considered to be infinite and the boundary conditions
are relatively simple. To deal with complex geometries and boundary conditions, an
accurate and efficient numerical method is essential. The finite difference method [61],
finite element method [108] and boundary element method [18, 20] are well known
numerical methods for solving fracture mechanics problems.
1.4.1 Finite Difference Method
The finite difference method solves the differential field equations by considering dis-
crete values of the mesh points throughout the domain of interest. This results in
a large system of algebraic equations, which require the use of considerable compu-
tational power for calculating the solution. To obtain accurate results for the finite
difference method, a good understanding of the effects of discretisation is important.
The finite difference method is often used in applications that require very fine meshes
and a large number of repeated operations because of the straightforward procedures
in matrix generation and manipulation.
1.4.2 Finite Element Method
The finite element method involves the approximation of the variables over the do-
main, in terms of polynomial interpolation functions. The domain is divided into
Chapter 1: Introduction 9
regions, called elements. These elements are interconnected at specified joints that
are called nodes or nodal points. These interpolation functions are defined in terms
of the values of the field variable at the nodal points. When the governing equa-
tions in finite element method are integrated over each finite element, the unknowns
will be the nodal values of the field variable. Continuity over adjacent elements is
applied. Assembling all of these together, a banded matrix is produced by solving
the field equations. The nodal value of the field variable can be obtained through-
out the assemblage of elements. Because each mesh is easily graded and a general
type of boundary conditions is incorporated, the finite element method is much more
powerful than the finite difference method. However, the main disadvantage of the
finite element method is that large quantities of data are still required because the
full domain needs to be discretised.
1.4.3 Boundary Element Method
In boundary value problems, an integral equation within a specified domain can be
transformed into a boundary integral equation over the boundary of the domain. In
order to solve the boundary integral equation numerically, the boundary of the domain
must be sub-divided into segments, known as boundary elements. The boundary
element method is based on Betti’s reciprocal theorem [84]. Assuming that (1) is the
loading of interest and (2) is a reference loading with a known solution, then in the
absence of body forces, Betti’s theorem can be stated as follows [11]:∫
S
T(1)i u
(2)i dS =
∫
S
T(2)i u
(1)i dS (1.2)
where S is the boundary of the domain, Ti and ui are components of the traction and
displacement vectors, respectively, and superscripts denote loadings (1) and (2).
Chapter 1: Introduction 10
Applying Betti’s reciprocal theorem to the boundary conditions, where a unit
force is applied at an interior point X′, resulting in displacements and tractions at
surface point x, leads to
ui(X′) = −
∫
S
Tij(X′,x)uj(x)dS(x) +
∫
S
Uij(X′,x)tj(x)dS(x) (1.3)
where Tij and Uij are the fundamental solutions, ui(X′) is the displacement vector
at the interior point X′, uj(x) and tj(x) are the reference displacement and traction
vectors at the boundary point x. If we let X′ → x′, where x′ is a boundary point,
equation (1.3) becomes:
1
2ui(x
′) +
∫
S
Tij(x′,x)uj(x)dS(x) =
∫
S
Uij(x′,x)tj(x)dS(x) (1.4)
assuming the boundary is smooth. In order to solve for the unknown boundary data,
the boundaries must be divided into elements, and equation (1.4) approximated by a
system of algebraic equations.
The boundary element method has emerged as a powerful alternative to the finite
element method. The most important features of the boundary element method is
that it reduces the dimensionality of the problem by one, resulting in a smaller system
of equations and a considerable reduction in the data required for the analysis. Hence,
the boundary element method programs are easier to use with existing solid modellers
and mesh generators. The advantage is particularly important for problems which
involve many modifications. Meshes can easily be generated and do not require a
complete re-meshing.
Chapter 1: Introduction 11
1.5 Advanced Formulations in Boundary Element
Method
The boundary element method is one of the most powerful numerical techniques for
the solution of problems in fracture mechanics. While much research has been done
in this area, more accurate and reliable solutions are still worthy to be explored. This
thesis presents some new boundary element formulations which have applications to
a wide range of problems.
1.5.1 Dual Boundary Element Method
The boundary element formulation in elastostatics is based on the displacement
boundary integral equation. It has been successfully applied to linear elastic prob-
lems in domains containing no degenerated area or volume. However, the solution of
general crack problems cannot be achieved in a single-region analysis with the direct
application of the boundary element method, because the coincidence of the crack
surfaces gives rise to a singular system of algebraic equations. The dual boundary
element method is formulated for the analysis of crack problems. The dual equations
are the displacement and the traction boundary integral equations. When the dual
boundary integral equations are applied, the displacement equation on one of the
crack surfaces and the traction equation on the other, general crack problems can be
solved in a single-region formulation.
In this thesis, the formulations for two dimensional linear elastic crack problems
under in-plane tensile and anti-plane shear loading, respectively, are proposed. Dis-
continuous quarter point element method and special crack tip element method are
Chapter 1: Introduction 12
used to correctly model the displacement in the vicinity of crack tips. The unknowns
on the crack surfaces are the relative displacement between upper and lower crack
surfaces. Once the relative displacements are solved numerically, physical quantities
of interest, such as crack tip stress intensity factors can be easily obtained.
1.5.2 Subregion Boundary Element Method
The need for more accurate and reliable solutions requires more computationally
intensive processes. The implementation of parallel processing techniques is one of
the best economical alternatives because of the advances of multiprocessor computing
technology.
In the last decade many attempts have been made to implement boundary ele-
ment method on parallel processing computers for the solution of various structural
mechanics problems and design problems. Domain decomposition method is very
useful for parallel processing in boundary element method. The method splits the
original, complex domain into smaller, simpler sub-domains assigned to different pro-
cessors. Thereafter the solution of the original problem can be reconstructed through
the solutions of these sub-problems. There are various techniques of domain decom-
position such as the iterative sub-structuring scheme, which uses domain partitions
splitting the original domain into small sub-domains and reduces the original problem
to an interface problem solved by an iterative method. The execution time in parallel
processing is determined by the processor with the largest execution time.
This thesis is concerned with a new subregion boundary element technique, in
which the system of equations for each subregion is transformed into an appropriate
form independently to utilize parallel computing. This technique is more efficient
Chapter 1: Introduction 13
than traditional methods because it significantly reduces the size of the resulting
matrix, which in turn reduces the required computation time.
1.5.3 Dual Reciprocity Boundary Element Method
The main advantage of the boundary element method is its ability to provide a com-
plete problem solution in terms of boundary values only, which provides substantial
savings in computer time and data preparation. However, non-homogeneous terms
accounting for efforts such as distributed loads were included in the formulation by
means of domain integrals, thus making the boundary element technique lose the
attraction of its “boundary only” character. Many different approaches have been
developed to overcome this problem, such as analytical integration of the domain
integrals, the use of Fourier expansions, the Galerkin vector technique, the multiple
reciprocity method, and the dual reciprocity boundary element method. The dual
reciprocity boundary element method is the only general technique other than cell
integration. It is an accurate, straightforward, and more powerful method. This
technique transforms the domain integrals to the boundary, with the aid of the re-
ciprocal theorem applied for the second time, the first time being when formulating
problem in integral form. Thus, the dual reciprocity boundary element method could
be employed to preserve the dimensionality reduction advantage of boundary ele-
ment method for non-homogeneous problems. It can be used to solve non-linear and
time-dependent problems as well as problems with internal source distributions.
In this thesis, the dual reciprocity boundary element method employing the step
by step time integration technique is developed to analyse two dimensional dynamic
crack problems. The equation of motion is expressed in a boundary integral form using
Chapter 1: Introduction 14
elastostatic fundamental solutions. The dual reciprocity boundary element method is
combined with an efficient subregion boundary element method to overcome the dif-
ficulty of a singular system of algebraic equations in crack problems. Dynamic stress
intensity factors are calculated using the discontinuous quarter point elements.
Chapter 2
Dual Boundary Integral
Formulation for Two Dimensional
Crack Problems
An efficient integral equation formulation for two-dimensional linear elastic crack
problems is proposed with the displacement equation being used on the outer bound-
ary and the traction equation being used on one of the crack faces. Discontinuous
quarter point elements are used to correctly model the displacement in the vicinity of
crack tips. Using this formulation a general crack problem can be solved in a single-
region formulation, and only one of the crack faces needs to be discretised. Once
the relative displacements of the cracks are solved numerically, physical quantities of
interest, such as crack tip stress intensity factors can be easily obtained. Numeri-
cal examples are provided to demonstrate the accuracy and efficiency of the present
formulation.
15
Chapter 2: Dual Boundary Integral Formulation 16
2.1 Introduction
Fracture mechanics analysis plays a central role in structural integrity assessments.
Many engineering materials have various kinds of defects such as scratches and micro-
cracks which cannot be precluded. These defects tend to intensify the local stress
field and hence reduce the load bearing capacity of a component. At the same time
increasing demands for energy and material conservation requires that structures
be designed with smaller safety margins. Therefore, the study of events of existing
cracks has become increasingly important. The complicated nature of multi-crack
problems means that any meaningful solution has to be obtained numerically. The
need of developing an accurate and effective numerical technique in fracture mechanics
analysis is obvious.
Various numerical techniques have been used by researchers in the fracture me-
chanics community. Among them, the boundary element method is now widely re-
garded as the most accurate numerical tool for the analysis of crack problems in linear
elastic fracture mechanics (see Brebbia & Dominguez [18]). However, the solution of
a general crack problem cannot be achieved in a single region analysis by a direct
application of the boundary element method; the boundary integral equations for
two geometrically coincident points on both surfaces of a crack are identical, thus
resulting in a singular system of algebraic equations.
Some special techniques have been proposed to overcome this difficulty. Among
these are the crack Green’s function method [90], the displacement discontinuity
method [31], the multi-domain formulation [16] and dual boundary element method
(DBEM) [27, 25, 29, 80]. Comparing these methods, DBEM has certain apparent
advantages over others in that crack problems can be solved with a single region
Chapter 2: Dual Boundary Integral Formulation 17
formulation [10, 28, 99, 100]. However, the DBEM formulation employed by Portela et
al. [80] incorporates two independent boundary integral equations; the displacement
equation to model one of the crack surfaces, and the traction equation to model the
other. Since the former equation is also applied to the outer boundary, this approach
results in a large system of equations, and could therefore be time-consuming in
computation, especially for structures with multi-cracks.
Chen & Chen [25] proposed a different DBEM formulation; the displacement
integral equation is applied only on the outer boundary, and the traction integral
equation on one of the crack surfaces. In their formulation, relative displacement
of crack surfaces is introduced instead of the displacement. This has reduced the
total number of degrees of freedom and computational effort. However, constant
elements are used to discretise the crack surface; a virtual boundary connected to
one of the crack surfaces is needed to construct a closed integral path to evaluate the
hypersingular integral. Although good accuracy is reported for the examples given in
[25], to model cracks of arbitrary geometry, quadratic elements are usually necessary.
Also, more elements are needed to yield accurate results, which could lead to a loss
of the advantage in the DBEM.
In this chapter, we propose an integral equation formulation that combines Chen
& Chen’s approach with the crack modelling strategy of Portela et al. with some
modification of the latter by placing discontinuous quarter-point elements at crack
tips. In this formulation, the displacement equation is used on the outer boundary
only while the traction equation is used on one of the crack surfaces. The unknowns
are the relative displacement of the crack faces and the displacement of the outer
Chapter 2: Dual Boundary Integral Formulation 18
boundary. Discontinuous quadratic elements are used to model the relative displace-
ment. These elements not only satisfy the requirement of continuity of both tractions
and strains, but also lead to an easy implementation of collocation at crack tips, crack
kinks and crack-edge corners [80]. With the use of quadratic elements, the number
of elements needed for the crack face is small, which in turn reduces the size of the
resulting system of algebraic equations. Instead of using the more complicated J-
integral method as in Portela et al. [80], the stress intensity factor (SIF) is calculated
by using the near tip relative displacement. More accurate results are obtained by
placing discontinuous quarter-point elements at crack tips, which correctly model the
behaviour of the crack tip displacement. Above all, the present formulation can be
effectively used to study structures with arbitrary number and distribution of cracks,
especially edge cracks and cracks of arbitrary shapes due to the adoption of discon-
tinuous quadratic elements and the relative displacement on crack faces. Numerical
examples are provided to demonstrate the accuracy and efficiency of the present for-
mulation.
2.2 The Dual Boundary Integral Formulation
A finite two dimensional body enclosed by a boundary Γ is considered. The boundary
integral representation of the displacement ui at an internal point X′ is given by (the
body force term is neglected) [32]
ui(X′) +
∫
Γ
Tij(X′,x)uj(x)dΓ(x) =
∫
Γ
Uij(X′,x)tj(x)dΓ(x) (2.1)
where Uij and Tij (i, j = 1, 2) represent the Kelvin displacement and traction fun-
damental solutions respectively, at the boundary integration point x. The Kelvin
Chapter 2: Dual Boundary Integral Formulation 19
fundamental solutions are given as
Uij(X′,x) =
1
8πµ(1 − ν)
[
(3 − 4ν) ln
(
1
r
)
δij +∂r
∂xi
∂r
∂xj
]
(2.2)
and
Tij(X′,x) = − 1
4π(1 − ν)r
∂r
∂n
[
(1 − 2ν)δij + 2∂r
∂xi
∂r
∂xj
]
−(1 − 2ν)
(
∂r
∂xi
nj −∂r
∂xj
ni
)
(2.3)
where r is the distance between X′ and x; µ, ν and δij represent the shear modulus,
Poisson’s ratio and the Kronecker delta, respectively; n denotes the unit outward
normal vector at the point x on the boundary; ni and nj are the direction cosines of
the normal with respect to xi and xj.
In the absence of body forces, consider a homogeneous, isotropic linear elastic
body, occupying the domain Ω with N cracks, and enclosed by the boundary ΓS.
Equation (2.1) can be written as
ui(X′) +
∫
ΓS
Tij(X′,x)uj(x)dΓ(x) +
N∑
n=1
∫
Γ+n
Tij(X′,x+)uj(x
+)dΓ(x)
−N∑
n=1
∫
−Γ−
n
Tij(X′,x−)uj(x
−)dΓ(x) =
∫
ΓS
Uij(X′,x)tj(x)dΓ(x) (2.4)
+N∑
n=1
∫
Γ+n
Uij(X′,x+)tj(x
+)dΓ(x) −N∑
n=1
∫
−Γ−
n
Uij(X′,x−)tj(x
−)dΓ(x)
where x+ and x− are the boundary points on the upper and lower crack surfaces,
respectively, ΓS represents the outer boundary, Γ+n the nth upper crack boundary,
Γ−n the nth lower crack boundary, and Γ = ΓS +
∑Nn=1(Γ
+n + Γ−
n ). Using the facts
that Uij(X′,x+)|Γ+ = Uij(X
′,x−)|−Γ− and Tij(X′,x+)|Γ+ = Tij(X
′,x−)|−Γ− , equation
Chapter 2: Dual Boundary Integral Formulation 20
(2.4) becomes
ui(X′) +
∫
ΓS
Tij(X′,x)uj(x)dΓ(x) +
N∑
n=1
∫
Γ+n
Tij(X′,x+)∆uj(x)dΓ(x)
=
∫
ΓS
Uij(X′,x)tj(x)dΓ(x) +
N∑
n=1
∫
Γ+n
Uij(X,x+)∆tj(x)dΓ(x) (2.5)
where ∆uj = uj(x+) − uj(x
−) and ∆tj = tj(x+) − tj(x
−). However, ∆tj is always
zero on the crack faces. Utilizing this fact and moving the source point X′ to the
outer boundary, we obtain the displacement equation [25]
cij(x′)uj(x
′) +
∫
ΓS
− Tij(x′,x)uj(x)dΓ(x) +
N∑
n=1
∫
Γ+n
Tij(x′,x+)∆uj(x)dΓ(x)
=
∫
ΓS
Uij(x′,x)tj(x)dΓ(x) (2.6)
where∫
− represents for the Cauchy principal value integral, cij(x′) is given by δij/2
for a smooth boundary at the point x′, and δij is the Kronecker delta.
In the absence of body forces, the stress components are obtained by differentiating
equation (2.5), followed by the application of Hooke’s law. The stress components
σij, at an internal point X′, can be expressed as
σij(X′) +
∫
ΓS
Skij(X′,x)uk(x)dΓ(x) +
N∑
n=1
∫
Γ+n
Skij(X′,x+)∆uk(x)dΓ(x)
=
∫
ΓS
Dkij(X′,x)tk(x)dΓ(x) (2.7)
where
Skij(X′,x) =
2µ
r2
2∂r
∂n
[
(1 − 2ν)δij∂r
∂xk
+ ν
(
δik∂r
∂xj
+ δjk∂r
∂xi
)
− 4∂r
∂xi
∂r
∂xj
∂r
∂xk
]
+ 2ν
(
ni∂r
∂xj
∂r
∂xk
+ nj∂r
∂xi
∂r
∂xk
)
+ (1 − 2ν)
(
2nk∂r
∂xi
∂r
∂xj
+ njδik + niδjk
)
− (1 − 4ν)nkδij
1
4π(1 − ν)(2.8)
Chapter 2: Dual Boundary Integral Formulation 21
and
Dkij(X′,x) =
1
r
(1 − 2ν)
[
δki∂r
∂xj
+ δkj∂r
∂xi
− δij∂r
∂xk
]
+ 2∂r
∂xi
∂r
∂xj
∂r
∂xk
1
4π(1 − ν)
(2.9)
In equation (2.7), Skij and Dkij are linear combinations of the derivatives of Tij and
Uij, respectively. Again, by moving the source point X′ to the upper crack boundary
x′, and using tj = σijni, where ni denotes the ith component of the outward normal
to the boundary at point x′ , we obtain the traction integral equation
1
2tj(x
′) + ni(x′)
∫
ΓS
Skij(x′,x)uk(x)dΓ(x) + ni(x
′)N∑
n=1
∫
Γ+n
= Skij(x′,x+)∆uk(x)dΓ(x)
= ni(x′)
∫
ΓS
Dkij(x′,x)tk(x)dΓ(x) (2.10)
where∫
= represents the Hadamard [50] principal value integral. Both Cauchy and
Hadamard principal-value integrals in equations (2.6) and (2.10) are finite parts of
improper integrals [60, 62]. The treatment of these finite part integrals follows the
method in Portela et al. [80].
Equations (2.6) and (2.10) are the governing equations to be solved in terms of the
displacement on the outer boundary and the relative displacement on the crack faces.
Equation (2.6) is applied for collocation on the outer boundary where continuous
quadratic elements are used, whereas equation (2.10) is applied on the upper crack
faces which are modelled by discontinuous quadratic elements.
By taking all the discretised nodes on the outer boundary ΓS and upper crack
surfaces∑N
n=1 Γ+n as the source point x′, the system of equations (2.6) and (2.10) for
the multiple cracks problem can be written in a matrix form as
[
T1 T2 0
S1 S2 I
]
uS
∆uc
t+c
=
[
U1
D1
]
[tS] (2.11)
Chapter 2: Dual Boundary Integral Formulation 22
where T1, T2, U1 and S1, S2, D1 are the corresponding assembled matrices from
equations (2.6) and (2.10). uS is the displacement vector and tS is the traction
vector on the outer boundary ΓS. ∆uc is the relative displacement vector of the
crack surfaces and t+c is the traction vector on the upper crack face.
2.3 Stress Intensity Factor Calculation
The ultimate task in fracture mechanics analysis is the calculation of the stress inten-
sity factor which is a local parameter. The most common methods of evaluation are
the J-integral method and the near tip displacement method. The later is much pre-
ferred computationally since the calculation is straight forward. However, to obtain
accurate results, the singular nature of the crack tip displacement has to be modelled
correctly. The required singularity can be achieved by placing special elements at
crack tips [44, 71]. Discontinuous quarter-point crack tip elements are used in the
present formulation. The stress intensity factors are calculated as
KI,II =µ
κ + 1
√
2π
r∆un,t(r) (2.12)
where κ = 3 − 4ν for plane strain problems, r is the distance from crack tip to the
nearest node on the upper crack face. ∆un(r) and ∆ut(r) denote the relative normal
and tangential displacement at r. The results show that the inclusion of the special
crack tip elements led to improved accuracy and efficiency in the SIF calculation.
Chapter 2: Dual Boundary Integral Formulation 23
2.4 Numerical Examples
To demonstrate the accuracy and efficiency of the present formulation, three cases are
considered, where a finite plate with different crack profiles is subjected to uniform
tension in one direction. In all the calculations, plane strain is assumed with a
Poisson’s ratio ν = 0.3.
First, consider a rectangular plate with a central slant crack subject to a uniform
traction T as shown in Figure 2.1. This is an example from Portela et al. [80].
Figure 2.1: Rectangular plate with a central slant crack (h/w = 2, θ = 45)
For the comparison purpose, the parameters are kept the same here. In solving this
problem, 30 quadratic elements were used compared to 36 in Portela et al. [80].
The stress intensity factors were calculated using the near tip relative displacements,
which requires very little calculation, while 30 internal points were needed in the
Chapter 2: Dual Boundary Integral Formulation 24 J – integral approach [80] for a similar degree of accuracy. As can be seen in Figures 2.2 and 2.3,
the present results match those of References [73] and [80] within two decimal places.
Figure 2.2: Normalised mode I SIF for the rectangular plate with a central slant crack: (a) the present method, (b) Reference [80], and (c) Reference [73]
Please see print copy for Figure 2.2
Chapter 2: Dual Boundary Integral Formulation 25
Figure 2.3: Normalised mode II SIF for the rectangular plate with a central slant crack: (a) the present method, (b) Reference [80], and [c] Reference [73]
Please see print copy for Figure 2.3
Chapter 2: Dual Boundary Integral Formulation 26 For a/w = 0.1, the results of normalized mode I and mode II SIF for different θ are plotted in
Figures 2.4 and 2.5, respectively. The results obtained are remarkably accurate with the difference
between the present and the Reference [73] within three decimal places.
Figure 2.3: Normalised mode II SIF for the rectangular plate with a central slant crack: (a) the present method, (b) Reference [80], and [c] Reference [73]
Please see print copy for Figure 2.4
Chapter 2: Dual Boundary Integral Formulation 27
Figure 2.5: Normalised mode II SIF for the rectangular plate with a central slant crack (a/w = 0.1): (a) the present method, and (b) Reference [73]
Please see print copy for Figure 2.5
Chapter 2: Dual Boundary Integral Formulation 28
The second example is a rectangular plate with an internal kinked crack as shown
in Figure 2.6. The loading condition and other physical characters are again kept
Figure 2.6: Rectangular plate with an internal kinked crack (h/w = 2, a/w = 0.1)
the same as in Portela et al. [80]. The stress intensity factors at both tips A and
B were calculated and compared with the results published by Murakami [73] and
those obtained by the J−integral approach [80]. Again, excellent agreement is found
among the results as shown in Tables (2.1–2.2), where 2c = a +√
22
b.
Chapter 2: Dual Boundary Integral Formulation 29
Table 2.1: Mode I SIF for an internal kinked crack in a rectangular plate
KI/T√
πctip A tip B
b/a 0.2 0.4 0.6 0.2 0.4 0.6Present 0.996 0.991 0.988 0.600 0.575 0.569
Reference [73] 0.995 0.990 0.986 0.598 0.574 0.568Reference [80] 0.993 0.989 0.987 0.604 0.576 0.570
Table 2.2: Mode II SIF for an internal kinked crack in a rectangular plate
KII/T√
πctip A tip B
b/a 0.2 0.4 0.6 0.2 0.4 0.6Present 0.027 0.034 0.031 0.568 0.609 0.630
Reference [73] 0.028 0.033 0.030 0.557 0.607 0.627Reference [80] 0.030 0.036 0.032 0.556 0.603 0.624
The fracture condition for mixed mode is obtained by [9]
K = KI cos3 φ
2− 3KII cos2 φ
2sin
φ
2(2.13)
where
φ = 2 tan−1
1
4
KI
KII
± 1
4
√
(
KI
KII
)2
+ 8
While K reaches the critical value of the stress intensity factor, the crack will grow.
Using the results in Tables (2.1–2.2) we find that tip B is the worse case.
Chapter 2: Dual Boundary Integral Formulation 30
The following figures show the effect of crack ratio b/a on the stress intensity
factor. Different angle cases of the kinked crack are plotted. The results are shown
in Figures 2.7 and 2.8 for tip A, and Figures 2.9 and 2.10 for tip B.
b/a
Nor
mal
ised
mod
eI
SIF
attip
A
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10.95
0.96
0.97
0.98
0.99
1
1.01
1.02
1.03
1.04
1.05(a)(b)(c)
Figure 2.7: Normalised mode I SIF vs. crack ratio b/a at tip A. The angle of thekinked crack is (a) 30, (b) 45, (c) 60
Chapter 2: Dual Boundary Integral Formulation 31
b/a
Nor
mal
ised
mod
eII
SIF
attip
A
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-0.05
-0.04
-0.03
-0.02
-0.01
0
0.01
0.02
0.03
0.04
0.05(a)(b)(c)
Figure 2.8: Normalised mode II SIF vs. crack ratio b/a at tip A. The angle of thekinked crack is (a) 30, (b) 45, (c) 60
Chapter 2: Dual Boundary Integral Formulation 32
b/a
Nor
mal
ised
mod
eI
SIF
attip
B
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1(a)(b)(c)
Figure 2.9: Normalised mode I SIF vs. crack ratio b/a at tip B. The angle of thekinked crack is (a) 30, (b) 45, (c) 60
Chapter 2: Dual Boundary Integral Formulation 33
b/a
Nor
mal
ised
mod
eII
SIF
attip
B
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1(a)(b)(c)
Figure 2.10: Normalised mode II SIF vs. crack ratio b/a at tip B. The angle of thekinked crack is (a) 30, (b) 45, (c) 60
Chapter 2: Dual Boundary Integral Formulation 34
The third example is a finite plate with two inclined cracks as shown in Figure
2.11, where d/a = 1.1. It was discussed in Chen & Chang [24] and Chen & Chen
Figure 2.11: Rectangular plate with two inclined cracks (h/w = 2, a/W = 0.25)
[25], and all the physical parameters are kept the same for comparison. The outer
boundary is modelled with 20 quadratic elements and the crack surface is discretised
with only 6 discontinuous quadratic elements compared with 18 constant elements
in Chen & Chen [25]. Our results were compared with those of Chen & Chang [24],
because the numerical results in Chen & Chen [25] are unavailable, but were shown
to be in excellent agreement with Chen & Chang’s. As shown in Table 2.3, the
correlation between the two sets of results is excellent with the maximum difference
being less than 2 per cent. The results for an infinite plate were also calculated and
compared with analytical results [73] with a maximum error of about 0.1 per cent.
Chapter 2: Dual Boundary Integral Formulation 35
Table 2.3: Comparison of normalised stress intensity factors
KI,II/T√
πaPresent Reference [24]
mode I mode II mode I mode IIθ A B A B A B A B45 0.596 0.520 0.595 0.518 0.513 0.557 0.511 0.55260 0.981 0.846 0.976 0.834 0.487 0.516 0.482 0.51175 1.405 1.128 1.394 1.113 0.316 0.319 0.313 0.31490 1.636 1.247 1.628 1.227 0 0 0 0
Chapter 2: Dual Boundary Integral Formulation 36
Figures 2.12 and 2.13 show the effects of the interaction of the cracks on stress
intensity factors. Because of the interaction of the cracks, the stress intensity factors
at A are greater than those at B. The results for an infinite plate are also plotted in
the same figures to show the effect of finite size.
θ
Nor
mal
ised
mod
eI
SIF
0 10 20 30 40 50 60 70 80 900
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8(a)(b)(c)(d)
Figure 2.12: Normalised mode I SIF vs. inclined angle θ: (a) Finite region (b) Infiniteregion at tip A, and (c) Finite region (d) Infinite region at tip B
Chapter 2: Dual Boundary Integral Formulation 37
θ
Nor
mal
ised
mod
eII
SIF
0 10 20 30 40 50 60 70 80 900
0.1
0.2
0.3
0.4
0.5
0.6(a)(b)(c)(d)
Figure 2.13: Normalised mode II SIF vs. inclined angle θ: (a) Finite region (b) Infiniteregion at tip A, and (c) Finite region (d) Infinite region at tip B
Chapter 2: Dual Boundary Integral Formulation 38
The final example is a rectangular plate with two parallel cracks, as shown in Fig-
ure 2.14. The continuous and discontinuous quadratic elements are used to discretise
Figure 2.14: Rectangular plate with two parallel cracks (h/w = 2, a/w = 0.025)
the outer boundary and the crack faces, respectively. There are 36 elements used, in
which there are 6 elements on each crack face. The normalised mode I SIF is plotted
in Figure 2.15, where s = a/(a+d). The results agree well with the published results
[93] with a maximum difference of about 1.5 per cent. Also, the results at different
ratio a/w are plotted on figures 2.16 and 2.17.
Chapter 2: Dual Boundary Integral Formulation 39
Figure 2.15: Normalised mode I SIF for the rectangular plate with two parallel cracks: (a) the present method, and (b) Reference [93]
Please see print copy for Figure 2.15
Chapter 2: Dual Boundary Integral Formulation 40
s
Nor
mal
ised
mod
eI
SIF
0.1 0.2 0.3 0.4 0.50.8
0.85
0.9
0.95
1
1.05
1.1(a)(b)(c)(d)
Figure 2.16: Normalised mode I SIF vs. s, with a/w given by (a) 0.025, (b) 0.05, (c)0.1, (d) 0.2
Chapter 2: Dual Boundary Integral Formulation 41
s
Nor
mal
ised
mod
eII
SIF
0.1 0.2 0.3 0.4 0.50
0.01
0.02
0.03
0.04
0.05
0.06
0.07(a)(b)(c)(d)
Figure 2.17: Normalised mode II SIF vs. s, with a/w given by (a) 0.025, (b) 0.05, (c)0.1, (d) 0.2
Chapter 2: Dual Boundary Integral Formulation 42
2.5 Conclusion
In this chapter, an efficient boundary integral equation formulation was presented
to solve linear elastic crack problems. The formulation incorporates two boundary
integral equations; the displacement equation on the outer boundary and the traction
equation on one of the crack faces. General crack problems can be solved in a single
region formulation. The formulation is specially useful in the analysis of cracks of
arbitrary geometry, as well as structures with multi-cracks. It has the advantage over
the existing dual boundary element methods on saving of computing time because
fewer elements are needed. The use of discontinuous quarter-point elements at the
crack tips correctly describes the r1/2 behaviour of the near tip displacement, and
therefore leads to increased accuracy and efficiency.
Chapter 3
An Efficient Dual Boundary
Element Method for Crack
Problems with Anti-plane Shear
Loading
This chapter is concerned with an efficient dual boundary element method for 2-d
crack problems under anti-plane shear loading. The dual equations are the displace-
ment and the traction boundary integral equations. When the displacement equation
is applied on the outer boundary and the traction equation on one of the crack sur-
faces, general crack problems with anti-plane shear loading can be solved with a sin-
gle region formulation. The outer boundary is discretised with continuous quadratic
elements, however, only one of the crack surfaces needs to be discretised with dis-
continuous quadratic elements. Highly accurate results are obtained, when the stress
intensity factor is evaluated with the discontinuous quarter point element method.
Numerical examples are provided to demonstrate the accuracy and efficiency of the
present formulation.
43
Chapter 3: Anti-plane Shear Loading 44
3.1 Introduction
The problem of a cracked body subjected to an anti-plane shear loading had been
studied extensively. Sih [87] provided analytical solutions for mode III cracks in
infinite regions by using Westergaard stress functions and Muskhelishvili’s method.
Chiang [26] presented analytical solutions for slightly curved cracks in anti-plane
strain in infinite regions using perturbation procedures similar to those carried out
for in-plane loading cases by Cotterell & Rice [30]. Zhang [105, 106] and Ma & Zhang
[70] gave analytical solutions for a mode III stress intensity factor considering a fi-
nite region with an eccentric straight crack. Ma [69] provided analytical solutions for
mode III straight cracks in finite regions using Fourier transforms and Fourier series.
However, their solutions were concerned with specified geometries or boundary condi-
tions. To deal with the complexities of general geometries and boundary conditions,
an accurate and efficient numerical method is essential [48, 52, 100, 107].
Several numerical solutions had been devised for anti-plane crack problems. Paulino
et al. [77] provided numerical solutions for a curved crack subjected to an anti-plane
shear loading in finite regions by using the boundary integral equation method. Ting
et al. [94] provided numerical solutions for mode III crack problems by using the
boundary element alternating method. Liu & Altiero [63] provided numerical solu-
tions for mode III crack problems using the boundary integral equation with linear
approximation on displacements and stresses. Mews & Kuhn [72] provided numerical
solutions for the traction free central crack problem by using Green’s function, instead
of the usual fundamental solution. Sadegh & Altiero [5] used the indirect boundary
integral equation method to solve traction problems, using displacement-based for-
mulations. In general, the boundary element method (BEM) is a well established
Chapter 3: Anti-plane Shear Loading 45
numerical technique for the analysis of linear fracture mechanics problems. However,
the solution of general crack problems cannot be achieved with the direct application
of the BEM, because the coincidence of the crack surfaces gives rise to a singular
system of algebraic equations.
To overcome this shortcoming, we provide an efficient numerical procedure, based
on the dual boundary element method (DBEM), for anti-plane shear loading prob-
lems. The dual boundary element method seems to have certain apparent advantages
for in-plane loading problems with a single region formulation [66]. This method incor-
porates two independent boundary integral equations, the displacement and traction
equations. Portela et al. [80] considered the effective numerical implementation of
the two dimensional DBEM for solving general in-plane fracture mechanics problems.
Chen & Chen [25] proposed a different DBEM formulation for in-plane crack prob-
lems. Chen & Chen suggested the use of the displacement integral equation applied
only on the outer boundary and the traction integral equation on one of the crack
surfaces. In Chen & Chen’s formulation, relative displacement of crack surfaces was
used instead of the displacement. This reduces the degrees of freedom and hence the
computational effort. This study uses an integral equation formulation that combines
Lu & Wu’s [66] approach with the crack modelling strategy of quadratic boundary
elements for anti-plane crack problems. The stress intensity factor is calculated based
on the near tip displacement method. More accurate results are obtained by placing
discontinuous quarter point elements at crack tips [85], which correctly model the
behaviour of the crack tip displacement. This is a similar technique to that used for
continuous quarter point elements [71]. Numerical examples are provided to demon-
strate the accuracy and efficiency of the present formulation.
Chapter 3: Anti-plane Shear Loading 46
3.2 The Dual Boundary Integral Equation
for Anti-plane Problems
Consider a finite domain subjected to an arbitrary anti-plane shear loading, where
the only nonzero displacement component uz in the z direction may be specified as
follows [69]:
∇2uz = 0 (3.1)
The Laplace equation (3.1) can be transformed into a boundary integral equation,
as is typical with the BEM. The boundary integral formulation of the displacement
component, uz, at an internal point X′, is given by [18]
uz(X′) +
∫
Γ
H(X′,x)uz(x)dΓ(x) =
∫
Γ
G(X′,x)tz(x)dΓ(x) (3.2)
where H(X′,x) and G(X′,x) represent the fundamental traction and displacement
solutions, respectively, at a boundary point x, which are given as
H(X′,x) = − 1
2πr
∂r
∂n(3.3)
and
G(X′,x) =1
2πµln
(
1
r
)
(3.4)
where µ is the shear modulus, r is the distance between X′ and x, and n denotes the
outward normal unit vector at the point x on the boundary Γ.
Chapter 3: Anti-plane Shear Loading 47
If we consider a finite body with L cracks, equation (3.2) can be written as
uz(X′) +
∫
ΓS
H(X′,x)uz(x)dΓ(x) +L∑
l=1
∫
Γ+
l
H(X′,x+)uz(x
+)dΓ(x)
+L∑
l=1
∫
Γ−
l
H(X′,x−)uz(x
−)dΓ(x)
=
∫
ΓS
G(X′,x)tz(x)dΓ(x) +L∑
l=1
∫
Γ+
l
G(X′,x+)tz(x
+)dΓ(x)
+L∑
l=1
∫
Γ−
l
G(X′,x−)tz(x
−)dΓ(x) (3.5)
where x+ and x− are the field points located on upper and lower crack surfaces,
respectively. Note that ΓS denotes the outer boundary of the body, Γ+l the lth upper
crack boundary, Γ−l the lth lower crack boundary, and Γ = ΓS+
∑Ll=1(Γ
+l +Γ−
l ). Using
the fact that H(X′, x+)|Γ+ = H(X′,x
−)|−Γ− and G(X′,x
+)|Γ+ = G(X′,x
−)|−Γ− ,
equation (3.5) can be simplified to
uz(X′) +
∫
ΓS
H(X′,x)uz(x)dΓ(x) +L∑
l=1
∫
Γ+
l
H(X′,x+)∆uz(x)dΓ(x)
=
∫
ΓS
G(X′,x)tz(x)dΓ(x) +L∑
l=1
∫
Γ+
l
G(X′,x+)∆tz(x)dΓ(x) (3.6)
where ∆uz = uz(x+)−uz(x
−) and ∆tz = tz(x+)− tz(x
−), however ∆tz is always zero
on the crack faces. As the internal point approaches the outer boundary, that is, as
X′ → x′, the displacement equation becomes
c(x′)uz(x′) +
∫
ΓS
− H(x′,x)uz(x)dΓ(x) +L∑
l=1
∫
Γ+
l
H(x′,x+)∆uz(x)dΓ(x)
=
∫
ΓS
G(x′,x)tz(x)dΓ(x) (3.7)
where∫
− represents the Cauchy principle value integral and c(x′) = 12, given a smooth
boundary at the point x′.
Chapter 3: Anti-plane Shear Loading 48
The stress components σiz are obtained from differentiation of equation (3.6),
followed by the application of Hooke’s law. At an internal point X′, these components
are given by
σiz(X′) +
∫
ΓS
Si(X′,x)uz(x)dΓ(x) +
L∑
l=1
∫
Γ+
l
Si(X′,x)∆uz(x)dΓ(x)
=
∫
ΓS
Di(X′,x)tz(x)dΓ(x) (3.8)
where Si(X′,x) and Di(X
′,x) are linear combinations of derivatives of H(X′,x) and
G(X′,x) in the i direction, which are given as
Si(X′,x) =
µ
2πr2
[
∂r
∂xi
∂r
∂n−(
δij −∂r
∂xj
∂r
∂xi
)
nj
]
(3.9)
and
Di(X′,x) = − 1
2πr
∂r
∂xi
(3.10)
where ni denotes the ith component of the outward normal to the boundary at point
x, and δij is the Kronecker delta. Again, by moving the source point X′ to the upper
crack boundary x′, and using tz = σizni, we obtain the traction integral equation
1
2tz(x
′) +
∫
ΓS
ni(x′)Si(x
′,x)uz(x)dΓ(x) +L∑
l=1
∫
Γ+
l
= ni(x′)Si(x
′,x)∆uz(x)dΓ(x)
=
∫
ΓS
ni(x′)Di(x
′,x)tz(x)dΓ(x) (3.11)
where∫
= represents the Hadamard principal value integral. Both Cauchy and Hadamard
principal-value integrals in equations (3.7) and (3.11) are finite parts of improper in-
tegrals. The treatment of the finite part integrals follows the method in Portela et
al. [80].
The displacement integral equation (3.7) and the traction integral equation (3.11)
are the governing equations to be solved for the displacement of the outer bound-
ary and the relative displacement of the crack faces. Equation (3.7) is applied for
Chapter 3: Anti-plane Shear Loading 49
collocation on the outer boundary where continuous quadratic elements are used,
whereas equation (3.11) is applied on the upper crack faces which are modelled by
discontinuous quadratic elements.
By taking all the discretised nodes on the outer boundary ΓS and upper crack
surfaces∑L
l=1 Γ+l as the source point x′, the system of equations (3.7) and (3.11) for
the multiple cracks problem can be written in a matrix form as
[
H1 H2 0
S1 S2 I
]
uz,S
∆uz,c
tz,c+
=
[
G1
D1
]
[tz,S] (3.12)
where H1, H2, G1 and S1, S2, D1 are the corresponding assembled matrices from
equations (3.7) and (3.11), respectively. The uz,S and tz,S are the displacement and
traction vectors on the outer boundary ΓS, respectively. ∆uz,c and tz,c+ are the
relative displacement vector and the traction vector on the upper crack faces.
3.3 Calculation of the Mode III Stress Intensity
Factor
Near tip displacement extrapolation is used to evaluate the numerical values of the
stress intensity factor. The relative displacements of the crack surfaces are calculated
using the DBEM and are used in the near crack tip stress field equations to obtain
the stress intensity factor. Due to the singular behaviour of the stress around the
crack tip, it is reasonable to expect a better approximation by replacing the normal
discontinuous quadratic element with a transition element possessing the same order
of singularity at the crack tip. The discontinuous quarter point element method
Chapter 3: Anti-plane Shear Loading 50
is used in the present formulation [42, 85]. The mode III stress intensity factor is
evaluated as:
KIII =µ
4
√
2π
r∆uz(r) (3.13)
where r is the distance from the crack tip to the nearest node on the upper crack
face, and ∆uz(r) denotes the relative displacement in the anti-plane direction.
3.4 Numerical Examples
In order to demonstrate the accuracy and efficiency of the technique previously de-
scribed, and to illustrate possible applications, we now consider several examples. In
all the numerical tests, the outer boundary is modelled by 24 continuous quadratic
elements, and each crack discretisation is carried out with three different meshes of
6, 8, and 10 discontinuous quadratic elements, respectively. The best accuracy is
achieved with 6 elements, in which the crack discretisation is graded, towards the tip,
with ratios 0.25, 0.15, and 0.1. The plate is subjected to a uniform anti-plane shear
loading τ , the stress intensity factor is normalised with respect to
K0 = τ√
πa (3.14)
where a defines the half length of the crack. All computations are carried out under
the condition of plane strain.
Firstly, consider a rectangular plate containing a central slant crack as shown
in Figure 3.1. The crack has length 2a and makes an angle θ with the horizontal
direction. For a horizontal crack (θ = 0), the normalised mode III stress intensity
factor is calculated for various ratios of a/h and a/w, and compared to those given
in References [63] and [68] (see Table 3.1). The largest difference between these does
Chapter 3: Anti-plane Shear Loading 51
Figure 3.1: Rectangular plate with a central slant crack
not exceed 1.65 per cent. Further, the normalised mode III stress intensity factor is
calculated for h/w = 2, while the crack is slanted an angle θ with the various ratios
of a/w. Three cases are considered, where θ = 30, 45, and 60, respectively. The
results obtained are presented in Figure 3.2. As it can be seen, when the ratio of a/w
increases, the stress intensity factor increases due to edge effect.
Chapter 3: Anti-plane Shear Loading 52
Table 3.1: Normalised mode III stress intensity factor for a straight central cracka : h 1 : 0.25 1 : 0.5 1 : 1 1 : 2 1 : 4
a : w Present 1.909 1.724 1.689 1.688 1.6611:1.2 Reference [63] 1.897 1.723 1.689 1.686 1.686
Reference [68] 1.900 1.725 1.691 1.689 1.689a : w Present 1.796 1.467 1.371 1.361 1.3611:1.4 Reference [63] 1.780 1.460 1.369 1.359 1.358
Reference [68] 1.782 1.463 1.370 1.361 1.360a : w Present 1.784 1.405 1.257 1.236 1.2361:1.6 Reference [63] 1.771 1.399 1.254 1.233 1.233
Reference [68] 1.773 1.401 1.256 1.235 1.235a : w Present 1.792 1.384 1.179 1.131 1.1291:2.0 Reference [63] 1.770 1.377 1.176 1.127 1.126
Reference [68] 1.772 1.379 1.178 1.130 1.128
Chapter 3: Anti-plane Shear Loading 53
a/w
Nor
mal
ised
mod
eII
IS
IF
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.80.5
0.6
0.7
0.8
0.9
1
1.1
1.2(a)(b)(c)
Figure 3.2: Normalised mode III stress intensity factor (SIF) for the rectangular platewith a central slant crack (a) θ = 30, (b) θ = 45, and (c) θ = 60
Chapter 3: Anti-plane Shear Loading 54
For the case where a/w = 1/50, which could be considered as the case of infinite
geometry since a << w, we compare the results with the analytical results for the
latter as given in Reference [93]. The results are plotted in Figure 3.3. Excellent
agreement is observed; the maximum error is around 0.02 per cent.
θ
Nor
mal
ised
mod
eII
IS
IF
0 10 20 30 40 50 60 70 80 900
0.2
0.4
0.6
0.8
1
1.2(a)(b)
Figure 3.3: Normalised mode III SIF for the infinite plate with a central slant crack(a) the analytical solutions, and (b) the present method
Chapter 3: Anti-plane Shear Loading 55
As shown in Figure 3.4, the second example is a rectangular plate containing
two identical collinear cracks. 2a is the length of the inclined crack and 2d is the
Figure 3.4: A finite plate with two collinear cracks
distance between the centre of the cracks. The geometric parameters are h/w = 2 and
a/w = 1/50. Figure 3.5 displays the variations of normalised mode III stress intensity
factors at tip A and tip B versus different ratios of a/d. Due to the interaction between
the two cracks, the computed normalised mode III stress intensity factor at tip A is
always larger then that at tip B. Hence, as the crack centre distance d decreases, the
difference of stress intensity factor increases. There is excellent correlation between
the computed results using the present method and those from analytical solutions;
the difference between these results does not exceed 0.03 per cent at tip A, or 0.09
per cent at tip B.
Chapter 3: Anti-plane Shear Loading 56
a/w
Nor
mal
ised
mod
eII
IS
IF
0.1 0.2 0.3 0.4 0.5 0.6 0.71
1.02
1.04
1.06
1.08
1.1
1.12
1.14(a)(b)(c)(d)
Figure 3.5: Normalised mode III SIF for the rectangular plate with two identicalcollinear cracks at tip A: (a) the analytical results, (b) the present method and attip B: (c) the analytical solutions, (d) the present method
Chapter 3: Anti-plane Shear Loading 57
The third example is an infinite plate (h/w = 2, a/w = 1/50) containing two
parallel cracks, as shown in Figure 3.6. 2a is the length of the two identical cracks
Figure 3.6: A finite plate with two parallel cracks
and 2d is the distance between the cracks. The computed results are compared with
the published results in Reference [93]. The results of normalised mode III stress
intensity factor for different s are plotted in Figure 3.7, where s = a/(a + d). The
effect of the interaction of cracks on the mode III stress intensity factor is observed.
The largest difference between the present and the published results does not exceed
0.65 per cent.
Chapter 3: Anti-plane Shear Loading 58
Figure 3.7: Normalised mode III SIF for the rectangular plate with two parallel cracks: (a) Reference [93] (b) the present method
Please see print copy for Figure 3.7
Chapter 3: Anti-plane Shear Loading 59
3.5 Conclusions
An efficient and accurate dual boundary element technique has been successfully es-
tablished for the analysis of two dimensional cracks subjected to an anti-plane shear
loading. The dual boundary equations are the usual displacement boundary integral
equation and the traction boundary integral equation. When the displacement equa-
tion is applied on the outer boundary and the traction equation is applied on one of
the crack surfaces, a general crack problem can be solved in a single region formu-
lation. The discontinuous quarter point elements are used for evaluating the mode
III stress intensity factor, which correctly describes the r1/2 behaviour of the near tip
displacements. This allows accurate results for mode III stress intensity factors to be
calculated.
Chapter 4
The Evaluation of Stress Intensity
Factors
The present chapter is concerned with developing a new technique for calculating the
stress intensity factor in two-dimensional linear elastic crack problems. This technique
is combined with the dual boundary element method to solve in-plane tensile loading
and anti-plane shear loading crack problems. Because quadratic boundary elements,
commonly used in boundary element method analysis, do not correctly describe the
behaviour of displacement near the crack tips, special crack tip elements are used to
model the displacement in the vicinity of crack tips. Numerical examples are provided
to demonstrate the accuracy, efficiency and stability of the present technique.
4.1 Introduction
The application of boundary element method (BEM) to fracture mechanics is now
well established and widely used in engineering. One of the main reasons is the
possibility of evaluating the stress intensity factors accurately and efficiently. The
stress intensity factors incorporate applied stress levels, geometry and crack size in a
60
Chapter 4: The Evaluation of Stress Intensity Factors 61
systematised manner and may be evaluated from the elastic stress analysis of cracked
structures. Irwin [57] showed that the stresses and displacements near the crack tip
could be described by a single constant that was related to the energy release rate.
This crack tip characterising parameter is known as the stress intensity factor. There
are three independent movements of the upper and lower crack surfaces with respect
to each other: (1) the crack opening mode (mode I), (2) the crack in-plane shear
mode (mode II) or sliding mode, and (3) the crack out-of-plane shear mode (mode
III) or tearing mode, as illustrated in Figure 1.1. The fracture in a cracked body
can be in any of these modes, or a combination of two or three modes.
Consider a crack problem in an infinite domain, Williams [102] had shown that
the stresses for an in-plane traction free crack may be written in terms of an infinite
series expansion with respect to the polar coordinates r and θ. Substituting Irwin’s
definition of the stress intensity factors into Williams’ expansion, we obtain the stress
components (see Figure 4.1) which are singular at the crack tip. As r → 0 the leading
Figure 4.1: Crack in an infinite plane
Chapter 4: The Evaluation of Stress Intensity Factors 62
terms in the stresses approach infinity, while the other terms remain finite or approach
zero. Consider only the first term of the Williams’ expression, the stress components
are expressed as [22]
σ11 =KI√2πr
cosθ
2
(
1 − sinθ
2sin
3θ
2
)
− KII√2πr
sinθ
2
(
2 + cosθ
2cos
3θ
2
)
(4.1)
σ22 =KI√2πr
cosθ
2
(
1 + sinθ
2sin
3θ
2
)
+KII√2πr
sinθ
2cos
θ
2cos
3θ
2(4.2)
σ12 =KI√2πr
sinθ
2cos
θ
2cos
3θ
2+
KII√2πr
cosθ
2
(
1 − sinθ
2sin
3θ
2
)
(4.3)
where KI and KII are the stress intensity factors corresponding to the opening mode
and the in-plane shear mode, respectively, and the size of r is much smaller than
the crack length. Integrating equations (4.1), (4.2), and (4.3) using the strain-
displacement and stress-strain relations, the displacement components in the vicinity
of the crack tip are
u1 =1
4µ
√
r
2π
KI
[
(2κ − 1) cosθ
2− cos
3θ
2
]
+KII
[
(2κ + 3) sinθ
2+ sin
3θ
2
]
(4.4)
u2 =1
4µ
√
r
2π
KI
[
(2κ − 1) sinθ
2− sin
3θ
2
]
−KII
[
(2κ − 3) cosθ
2+ cos
3θ
2
]
(4.5)
where µ is the shear modulus, κ = 3 − 4ν for plane strain, and ν the Poisson’s ratio.
Similarly, the anti-plane displacement component near the crack tip is expressed as
u3 = 4KIII
µ
√
r
2πsin
(
θ
2
)
(4.6)
where KIII is the stress intensity factor corresponding to the out-of-plane shear mode.
It is obvious from the above expressions that the stress and displacement fields vary
Chapter 4: The Evaluation of Stress Intensity Factors 63
with O(1/√
r) and O(√
r), respectively, in the vicinity of the crack tip, where r is the
distance to the crack tip.
4.2 The Numerical Evaluation of Stress Intensity
Factors
The standard BEM elements are continuous quadratic elements with the collocation
points at ξ = −1, 0, and 1. The interpolating polynomial (shape function) is
fi = φ1f 1i + φ2f 2
i + φ3f 3i (4.7)
where φ1 = 12ξ(ξ − 1), φ2 = 1 − ξ2, and φ3 = 1
2ξ(ξ + 1), see Figure 4.2.
Figure 4.2: General continuous quadratic element
The displacement and traction components may be approximated as
ui = a0 + a1
(r
l
)
+ a2
(r
l
)2
(4.8)
ti = b0 + b1
(r
l
)
+ b2
(r
l
)2
(4.9)
where aj and bj are constants, r denotes the distance to the crack tip and l denotes the
length of the element. Since the displacement and stress fields obtained do not contain
Chapter 4: The Evaluation of Stress Intensity Factors 64
O(1/√
r) and O(√
r), the standard continuous quadratic element does not model the
crack tip behaviours correctly and hence it may not lead to accurate results near
crack tip. There are several methods, such as the J-integral method [7, 9, 80, 81] or
the near tip displacement methods [16, 71, 74], to overcome this problem.
4.2.1 J-integral
The stress intensity factor can be related to a path independent integral, termed the
J-integral, described by Rice [81]. This integral is independent of a contour path
surrounding the crack tip chosen, as shown in Figure 4.3. The J-integral is defined
Figure 4.3: Coordinate reference system and contour path for J-integral
as
J =
∫
S
(
Wn1 − tj∂uj
∂n1
)
dS (4.10)
where S is an arbitrary contour surrounding the crack tip; W is the strain energy
density; tj are the traction components and n1 is the x1-component of the unit outward
Chapter 4: The Evaluation of Stress Intensity Factors 65
normal to the contour path. The relationship between the J-integral and the stress
intensity factors is given by [81]
J =K2
I + K2II
E ′(4.11)
where E ′ = E for plane stress conditions, and E ′ = E/(1 − ν2) for plane strain
conditions, while E is the elastic modulus. In order to decouple the stress intensity
factors in equation (4.11), we consider two points P (x1, x2) and P′
(x1,−x2) that are
symmetric relative to the axis along the crack surface, as shown in Figure 4.3. At
these points the displacement and stress fields can be expressed as a combination of
symmetric and antisymmetric components:
u1
u2
=
uI1 + uII
1
uI2 + uII
2
,
u′
1
u′
2
=
uI1 − uII
1
−uI2 + uII
2
(4.12)
and
σ11
σ22
σ12
=
σI11 + σII
11
σI22 + σII
22
σI12 + σII
12
,
σ′
11
σ′
22
σ′
12
=
σI11 − σII
11
σI22 − σII
22
−σI12 + σII
12
(4.13)
When equations (4.12) and (4.13) are substituted into equation (4.10), the J-integral
is represented by the sum of two integrals as follows [7]:
J = J I + J II (4.14)
where the superscripts indicate the deformation mode. Consequently, the J-integral
can be decoupled into mode I and mode II components, hence KI and KII can be
evaluated separately. Since the J-integral uses the potential energy theorem, the
elaborate representation of the crack tip singular fields is not necessary. However,
more source points surrounding the crack tip are required in the calculation of stress
intensity factors, because more points are needed for the contour integral calculation.
Chapter 4: The Evaluation of Stress Intensity Factors 66
4.2.2 The Discontinuous Quarter Point Element Method
The displacement and traction in a given crack problem can be expressed in terms
of singular and regular terms. The displacement and traction in the vicinity of the
crack tip are dominated by the singular terms which could be represented as unknown
stress intensity factors. The quarter point element methods [75, 79, 92, 103] are
used to model analytically the dominant√
r and 1/√
r behaviour exhibited by the
displacements and tractions in the vicinity of the crack tip. This can be achieved by
placing special elements at crack tips.
By shifting the midpoint node of a continuous quadratic element to the quarter
point position (close to the crack tip, see Figure 4.4), the displacement and traction
Figure 4.4: Singular quarter-point boundary elements
components are now expressed as follows
ui = A0 + A1
√
r
l+ A2
(r
l
)
(4.15)
ti = B0 + B1
√
r
l+ B2
(r
l
)
(4.16)
where Aj and Bj are constants. The displacement field in the vicinity of the crack is
correctly modelled to represent O(√
r) behaviour, but the traction in equation (4.16)
does not possess the correct 1/√
r singularity. However, in the boundary element
Chapter 4: The Evaluation of Stress Intensity Factors 67
method, the displacements and the tractions are represented independently. The
inclusion of the singularity can be accomplished by multiplying the right-hand side
of this equation by√
l/r [16], which then yields
ti =
√
l
r
[
B0 + B1
√
r
l+ B2
(r
l
)
]
= B0
√
l
r+ B1 + B2
√
r
l(4.17)
The above expression corresponds to the correct singular traction field O(1/√
r) in the
vicinity of the crack tip. Using the equations (4.15) and (4.17), both displacements
and tractions will be correctly represented.
In the boundary element method, subregion method is generally used in order to
avoid the problem that the coincidence of the crack surfaces gives rise to a singular
system of algebraic equations. The subregion method introduces artificial boundaries
into the cracked body, which connect the cracks to the boundaries. The prescribed
tractions and displacements along the crack surfaces are independently represented
in each domain. Then, the stress intensity factors can be computed by using the
equation (4.15) on the crack surface and equation (4.17) on the artificial boundary
around the crack tip elements. However, the selected artificial boundaries are not
unique and the method generates a large system of algebraic equations.
To overcome those shortcomings, the dual boundary element method was devel-
oped, which can solve the crack problem by a single region formulation. In this
method, both the displacement and the traction boundary integral equations are ap-
plied on the crack surfaces. An essential ingredient of the dual boundary element
formulation is the evaluation of the singular integrals. This feature requires a special
integration around the singular point (crack tip). Consider the collocation nodes at
the boundary elements, the displacement equation always satisfies the requirement of
the first order finite-part integrals (continuity of the displacement components at the
Chapter 4: The Evaluation of Stress Intensity Factors 68
nodes). However, the traction equation is required to satisfy the second order finite-
part integrals (continuity of the displacement derivatives at the nodes on a smooth
boundary). The discontinuous boundary elements implicitly provide the necessary
smoothness required, since the nodes are internal points of the element. Therefore,
the discretisation of the crack is best done with discontinuous boundary elements. In
addition, the problem of collocation at crack tips, crack kinks and crack-edge corners
is automatically circumvented by the use of discontinuous elements.
In order to keep the feature of the quarter point element method (shifting the
nodes to correctly model the behaviours of the displacement and traction on the crack
tip), we consider a discontinuous quadratic element, where the collocation points are
at ξ = −23, 0, and 2
3, respectively. By using the shape functions given in Figure 4.4,
the nodes of a discontinuous quadratic element are placed at r/l = 136
, 14, and 25
36,
respectively, where r denotes the distance to the crack tip and l denotes the length of
the element (see Figure 4.5). Thus, the displacement and traction variations for the
elements adjoining the crack tip are given by equations (4.15) and (4.17), respectively.
Figure 4.5: Modeling of the quarter point boundary element
Chapter 4: The Evaluation of Stress Intensity Factors 69
4.2.3 The Special Crack Tip Element Method
In this chapter, a particular method to evaluate the stress intensity factors is pre-
sented. Instead of using the interpolating polynomial functions on the near crack tip
element, we let the displacement be represented by
ui ≈ Bi
√r (4.18)
because of the O(√
r) behaviour of the near tip displacement field. The other bound-
aries are still discretised by using the standard interpolating polynomial functions.
Besides the unknown displacements of the collocation nodes, an unknown constant
Bi needs to be obtained. Let r and θ be a polar coordinate system with origin at the
crack tip, such that θ = ±π defines the crack faces. The relative displacement near a
traction free crack tip from equations (4.4), (4.5), and (4.6) can be written as
∆ux = ux(θ = π) − ux(θ = −π) =κ + 1
µKII
√
r
2π(4.19)
∆uy = uy(θ = π) − uy(θ = −π) =κ + 1
µKI
√
r
2π(4.20)
∆uz = uz(θ = π) − uz(θ = −π) =4
µKIII
√
r
2π(4.21)
where µ is the shear modulus, ν is the Poisson’s ratio and κ = 3−4ν for plane strain.
The stress intensity factors are given by
KI = Byµ
κ + 1
√2π (4.22)
KII = Bxµ
κ + 1
√2π (4.23)
KIII = Bzµ
4
√2π (4.24)
As mentioned in the previous section, in order to correctly model the behaviours of
the displacement and the traction in the vicinity of the crack tip, the collocation points
Chapter 4: The Evaluation of Stress Intensity Factors 70
of the near crack tip element need to be shifted to particular locations (quarter point
element). The locations of the nodal points depend on the choice of the quadratic
shape functions. On the contrary, the special crack tip element method is a relatively
straight-forward technique to follow, which correctly models the distribution of the
displacement and the traction in the near crack tip element.
4.3 Numerical Examples
In order to demonstrate the accuracy and efficiency of the proposed special crack tip
element method described in the previous section, the same examples as shown in
previous chapters, are calculated and the results are compared with those obtained
from other methods. All of the cases are under plane strain condition, and Poisson’s
ratio is taken to be ν = 0.3. The stress intensity factors are obtained from near
crack tip displacement calculations by using the discontinuous quarter point element
method (QPE) and the proposed special crack tip element method (SCT). The nu-
merical examples will show the results of the comparison of the stability and efficiency
between these methods. In each case the structure is subjected to a uniform traction
t or an anti-plane shear loading τ . The stress intensity factors are normalised with
respect to
K0 = t√
πa (4.25)
or
K0 = τ√
πa (4.26)
respectively, where a defines the half length of the crack.
Chapter 4: The Evaluation of Stress Intensity Factors 71
4.3.1 A Plate with a Central Slant Crack
When the crack length is small in comparison with plate dimensions (a/w = 140
and
h/w = 2), we could consider it as the case of an infinite region with a central crack.
So that we could compare our results with its analytical solutions. The geometry
of the present problem is shown in Figure 2.1 for in-plane problem or Figure 3.1 for
anti-plane problem. The analytical results for the stress intensity factors (SIF) are
given in Reference [93]:
KI = t cos2 θ√
πa (4.27)
KII = t cos θ sin θ√
πa (4.28)
KIII = τ cos θ√
πa (4.29)
The study is carried out with three different meshes of 6, 8, and 10 discontinuous
quadratic elements on each crack surface, respectively, and the outside boundary is
discretised with 24 quadratic elements. The crack discretisation is graded, towards
the tip, with the ratios 0.25, 0.15, and 0.1 for 6 elements, 0.2, 0.15, 0.1, and 0.0.5
for 8 elements, and equal size for 10 elements. The results obtained are shown in
Figures 4.6, 4.7 and 4.8. It can be seen that the special crack tip element method
is more stable than the discontinuous quarter point element method for different
crack discretisations. The results obtained with 6 crack elements show a higher level
of accuracy, the largest difference between the calculated results and the analytical
results does not exceed 0.08%.
Chapter 4: The Evaluation of Stress Intensity Factors 72
Figure 4.6: Relative error of Mode I SIF for the infinite plate with a central slantcrack from the QPE with (a) 6 elements, (b) 8 elements, (c) 10 elements and the SCTwith (d) 6 elements, (e) 8 elements, and (f) 10 elements
Chapter 4: The Evaluation of Stress Intensity Factors 73
Figure 4.7: Relative error of Mode II SIF for the infinite plate with a central slantcrack from the QPE with (a) 6 elements, (b) 8 elements, (c) 10 elements and the SCTwith (d) 6 elements, (e) 8 elements, and (f) 10 elements
Chapter 4: The Evaluation of Stress Intensity Factors 74
Figure 4.8: Relative error of Mode III SIF for the infinite plate with a central slantcrack from the QPE with (a) 6 elements, (b) 8 elements, (c) 10 elements and the SCTwith (d) 6 elements, (e) 8 elements, and (f) 10 elements
Chapter 4: The Evaluation of Stress Intensity Factors 75
The special crack tip element method can also produce comparably accurate re-
sults as that of the discontinuous quarter point element method for the finite region
problem. By using similar boundary discretisation, 24 quadratic element on the out-
side boundary and 6 on the crack face, three cases (a/w = 23, 1
2, 1
5) are studied, the
results are shown in Figures 4.9, and 4.10.
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
0 10 20 30 40 50 60 70 80 90
Mod
e I S
IF
Theta
QPE(a)(b)(c)
SCT(d)(e)(f)
Figure 4.9: Mode I SIF of the finite plate with a central slant crack from the QPEwith a
w= (a) 2
3, (b) 1
2, (c) 1
5and the SCT with (d) 2
3, (e) 1
2, (f) 1
5
Chapter 4: The Evaluation of Stress Intensity Factors 76
Figure 4.10: Mode II SIF of the finite plate with a central slant crack from the QPEwith a
w= (a) 2
3, (b) 1
2, (c) 1
5and the SCT with (d) 2
3, (e) 1
2, (f) 1
5
Chapter 4: The Evaluation of Stress Intensity Factors 77
4.3.2 Infinite Plate with Two Inclined Cracks
Consider, now, the analysis of two inclined cracks in an infinite plate (h/w = 2,
a/w = 140
). The plate is subjected to a normal loading t as shown in Figure 2.11 or
an anti-plane shear loading τ as in Figure 3.4. The crack has the length 2a and makes
an angle of θ = 90 with the vertical direction. The distance between the centre of
the cracks is 2d. The analytical results are given below [93]
KI,A =d − a
2a
√
d − a
d
[
(
d + a
d − a
)2E(k)
K(k)− 1
]
t√
πa (4.30)
KI,B =d + a
2a
√
d + a
d
(
1 − E(k)
K(k)
)
t√
πa (4.31)
for an in-plane normal loading and
KIII,A =d − a
2a
√
d − a
d
[
(
d + a
d − a
)2E(k)
K(k)− 1
]
τ√
πa (4.32)
KIII,B =d + a
2a
√
d + a
d
(
1 − E(k)
K(k)
)
τ√
πa (4.33)
for an anti-plane shear loading, where
k =
√
1 −(
d − a
d + a
)2
(4.34)
and
K(k) =
∫ π/2
0
1√
1 − k2 sin2 θdθ (4.35)
E(k) =
∫ π/2
0
√
1 − k2 sin2 θdθ (4.36)
To calculate the stress intensity factors with the discontinuous quarter point element
method and the special crack tip element method, the outside boundary was discre-
tised with 24 continuous quadratic elements and three different meshes of 6, 8, and
Chapter 4: The Evaluation of Stress Intensity Factors 78
10 discontinuous quadratic boundary elements were set up on each crack surface; the
elements are graded towards the tips, with the same ratios as mentioned in previous
example. The stress intensity factors were obtained for both tip A and tip B. The
error percentages for the methods are compared for the various crack discretisations,
the results are shown in Figures 4.11 and 4.12 for the mode I SIF and in Figures 4.13
and 4.14 for the mode III SIF. With such coarse meshes, the results obtained with
these methods are remarkably close; the results obtained by using the special crack
tip element method with 6 crack elements differ from the analytical results within
0.1% at tip A and 0.3% at tip B.
Figure 4.11: Relative error of Mode I SIF for the infinite plate with two inclinedcracks at tip A from the QPE with (a) 6 elements, (b) 8 elements, (c) 10 elementsand the SCT with (d) 6 elements, (e) 8 elements, and (f) 10 elements
Chapter 4: The Evaluation of Stress Intensity Factors 79
Figure 4.12: Relative error of Mode I SIF for the infinite plate with two inclinedcracks at tip B from the QPE with (a) 6 elements, (b) 8 elements, (c) 10 elementsand the SCT with (d) 6 elements, (e) 8 elements, and (f) 10 elements
Chapter 4: The Evaluation of Stress Intensity Factors 80
Figure 4.13: Relative error of Mode III SIF for the infinite plate with two inclinedcracks at tip A from the QPE with (a) 6 elements, (b) 8 elements, (c) 10 elementsand the SCT with (d) 6 elements, (e) 8 elements, and (f) 10 elements
Chapter 4: The Evaluation of Stress Intensity Factors 81
Figure 4.14: Relative error of Mode III SIF for the infinite plate with two inclinedcracks at tip B from the QPE with (a) 6 elements, (b) 8 elements, (c) 10 elementsand the SCT with (d) 6 elements, (e) 8 elements, and (f) 10 elements
Chapter 4: The Evaluation of Stress Intensity Factors 82
4.3.3 Infinite Plate with Two Parallel Cracks
As a final test, consider an infinite plate (h/w = 2, a/w = 140
) with two parallel cracks,
as shown in Figure 2.14 for in-plane problem or Figure 3.6 for anti-plane problem,
each crack has the length 2a and the distance between the cracks is 2d. The plate
is subjected to, at the ends, a uniform traction t for the in-plane problem or with
an anti-plane shear τ for the anti-plane problem. The stress intensity factors were
obtained by using three different boundary element meshes of quadratic elements as
mentioned in previous examples. The results are presented in Figures 4.15, 4.16, and
4.17. Comparing with the published results [93]
FI ≈ 1 − 0.293s[
1 − (1 − s)4]
(4.37)
and
FIII ≈ 1 − 0.293s2 (4.38)
where s = a/(a + d), our results showed excellent agreement with a maximum differ-
ence less than 1%; again, convergence was achieved with coarse meshes.
Chapter 4: The Evaluation of Stress Intensity Factors 83
0.84
0.86
0.88
0.9
0.92
0.94
0.96
0.98
1
1.02
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Nor
mal
ised
mod
e I S
IF
s
IsidaQPE(06)QPE(08)QPE(10)SCT(06)SCT(08)SCT(10)
Figure 4.15: Mode I SIF for the infinite plate with two parallel cracks from (a) theReference [93], the QPE with (b) 6 elements, (c) 8 elements, (d) 10 elements and theSCT with (e) 6 elements, (f) 8 elements, and (g) 10 elements
Chapter 4: The Evaluation of Stress Intensity Factors 84
0
0.01
0.02
0.03
0.04
0.05
0.06
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Nor
mal
ised
mod
e II
SIF
s
QPE(06)QPE(08)QPE(10)SCT(06)SCT(08)SCT(10)
Figure 4.16: Mode II SIF for the infinite plate with two parallel cracks from the QPEwith (a) 6 elements, (b) 8 elements, (c) 10 elements and the SCT with (d) 6 elements,(e) 8 elements, and (f) 10 elements
Chapter 4: The Evaluation of Stress Intensity Factors 85
0.92
0.93
0.94
0.95
0.96
0.97
0.98
0.99
1
1.01
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Nor
mal
ised
mod
e III
SIF
s
IsidaQPE(06)QPE(08)QPE(10)SCT(06)SCT(08)SCT(10)
Figure 4.17: Mode III SIF for the infinite plate with two parallel cracks from (a) theReference [93], the QPE with (b) 6 elements, (c) 8 elements, (d) 10 elements and theSCT with (e) 6 elements, (f) 8 elements, and (g) 10 elements
Chapter 4: The Evaluation of Stress Intensity Factors 86
4.4 Conclusion
Due to the O(1/√
r) and O(√
r) terms contained in the traction and displacement
fields, the standard interpolating polynomial functions can not accurately model dis-
placement and traction behaviours near crack tip. In order to correctly describe
the near crack tip behaviours, we presented the discontinuous quarter point element
method and the special crack tip element method for the evaluation of stress intensity
factors in two dimensional crack problems. In the discontinuous quarter point element
method, the locations of the collocation points needed to be chosen specially, which
depend on the shape functions at the crack tip element. However, in the special crack
tip element method it does not require shifting of the collocation points at the crack
tip element. This feature shows that it is easier to formalise the problem by using the
special crack tip element method. Furthermore, the special crack tip element method
slightly improves the performance because each crack tip element is treated as a sin-
gle unknown parameter only. Several examples indicate that the crack discretisation
has some effect on the accuracy of the numerical results of stress intensity factors.
Both methods obtained accurate results. The special crack tip element method has
a better stability than the discontinuous quarter point element method.
Chapter 5
A New Subregion Boundary
Element Technique
A new subregion boundary element technique is presented in this chapter. This
technique is applicable to the stress analysis of multi-region elastic media, such as
layered-materials. The technique is more efficient than traditional methods because
it significantly reduces the size of the final matrix. This is advantageous when a large
number of elements need to be used, such as in crack analysis. Also, as the system
of equations for each subregion is solved independently, parallel computing can be
utilized. Further, if the boundary conditions are changed the only equations required
to be recalculated are the ones related to the regions where the changes occur. This is
very useful for cases where crack extension is modelled with new boundary elements or
where crack faces come to contact. Numerical examples are presented to demonstrate
the accuracy and efficiency of the method.
87
Chapter 5: A New Subregion Boundary Element Technique 88
5.1 Introduction
Composite materials are increasingly used in various engineering structures, such as
in the aerospace and automotive industries. One of the advantages of these materials
is their ability to be tailored for individual applications. The use of composites could
be potentially limited by the lack of efficient methods to evaluate the strength and life
expectation of composite structures. While defects or micro-cracks are unavoidable,
they do have significant influence on the load transfer behaviour within the com-
posite. Due to the fact that composite materials are made of regions or zones with
different material properties, it is not always possible to utilise the general method for
homogenous materials. Therefore, it is crucial to develop accurate and efficient tech-
niques for numerical analysis of such materials, in case of fracture mechanics analysis,
calculating the stress intensity factor in layered materials with cracks.
A wide variety of analytical and numerical methods have been used to solve the
fracture problems of layered materials [13, 39, 53, 58, 82, 91]. If a straightforward
analytical solution is not possible, numerical procedures must be used. The finite
element method (FEM) is one of the most popular techniques to analyse fracture
problems in composite materials. The interior points have mesh connectivity to the
boundary points and extensive remeshing is required for crack propagation problems.
However, FEM remeshing for each crack length tends to be time consuming. In gen-
eral, the boundary element method (BEM) together with a subregion technique is
widely considered to be a very accurate numerical tool for the analysis of problems
where the materials consist of several homogeneous zones [18, 20]. All the bound-
aries of the body have to be discretised, including internal boundaries that separate
homogeneous zones. The BEM equations, constructed from all homogeneous zones
Chapter 5: A New Subregion Boundary Element Technique 89
combined with the interface traction and displacement continuity conditions, produce
a global matrix system. The numerical solution of this matrix system is the most
time consuming step of the numerical method, and hence can be the bottleneck for
the method being applied to problems that require a large number of elements.
Kita & Kamiya [59] presented a special method for the subregion boundary ele-
ment analysis to overcome this disadvantage. The linear system for each subregion
is transformed into equations similar to the stiffness equations of the FEM, and then
the global matrix equation is constructed by superposition of those equations for the
subregions. The matrix equation for each subregion is derived using the algorithm
in Brebbia & Georgiou [19]. This algorithm can be applied easily to objects divided
into subregions. The interface traction components are not obtained in the resulting
matrix system, but can be calculated from the equations for the subregions. The
technique has the advantage that the global coefficient matrix can be constructed
easily and a smaller system of algebraic equations is obtained. This method is more
effective for objects with multiple internal boundaries. However, a relatively large
global coefficient matrix is still needed.
Furthermore, for Kita & Kamiya’s method, in order to deduce the global matrix
system a matrix inversion for each subregion is required, which further increases com-
putation time. The number of numerical operations required for solving a system of
n linear equations and that in finding the inverse of a n × n matrix are of the same
order, n3, so even a slight increase in n increases computational time significantly.
Therefore, the reduction of computing time is an important task in practical cases.
High performance computing techniques, including parallel computing, are now being
applied in many engineering and scientific applications [86]. As a result, developing
Chapter 5: A New Subregion Boundary Element Technique 90
efficient numerical algorithms specifically aimed at high performance computers be-
comes a challenging issue.
In this chapter, an efficient technique for multi layer crack problems is proposed.
The matrices of all subregions are used to assemble the final interface traction ma-
trix, which is solved for the unknown interface traction components. Unlike other
methods which solve the displacement and traction components on the boundaries
and interfaces at the same time, the distribution of traction on the interfaces is ob-
tained first. The displacement components can then be calculated from the equations
associated with the corresponding subregions. Initially, extra numerical steps maybe
needed to set up the final interface matrix equation. However, our final matrix sys-
tem is significantly smaller than the final matrix systems obtained by other methods.
If the boundary conditions are changed, only the equations for the subregions con-
cerned need to be recalculated. Therefore, it greatly reduces computational time, and
provides overall efficiency.
The effects of crack size, layer size, and the material properties of the composite
on the stress intensity factor are studied using the proposed numerical technique to
demonstrate its accuracy and efficiency. The dual boundary element method (DBEM)
[25, 80, 66] is incorporated into the present method to overcome the singularity in
crack analysis. Further, in order to improve accuracy in the stress intensity factor
calculation, discontinuous quarter point elements [42, 85] are used to model the near
tip elements.
Chapter 5: A New Subregion Boundary Element Technique 91
5.2 The Multi Region Technique of Boundary El-
ement Method
Consider a two dimensional body consisting of several subregions. For any subre-
gion that contains no cracks, the displacement formulation of the boundary integral
equation, at a boundary point x′, is written in the form (the body force term is
neglected)
cij(x′)uj(x
′) +
∫
Γ
− Tij(x′,x)uj(x)dΓ(x) =
∫
Γ
Uij(x′,x)tj(x)dΓ(x) (5.1)
where∫
− stands for the Cauchy principal value integral. uj(x) and tj(x) are dis-
placement and traction components in the j direction, respectively. If the boundary
is smooth, cij(x′) = 1
2δij, where δij is the Kronecker delta. The kernel functions
Tij(x′,x) and Uij(x
′,x) represent the Kelvin traction and displacement fundamental
solutions, respectively, at the boundary point x. For any subregion containing cracks,
the DBEM is employed. The dual equations of the DBEM are the displacement and
the traction boundary integral equations. The traction equation, which is applied on
the crack surfaces, is obtained by differentiation of the displacement equation (5.1),
and followed by the application of Hooke’s law. It is written as
1
2tj(x
′) + ni(x′)
∫
Γ
= Skij(x′,x)uk(x)dΓ(x)
= ni(x′)
∫
Γ
− Dkij(x′,x)tk(x)dΓ(x) (5.2)
where∫
= stands for the Hadamard principal value integral, ni denotes the ith compo-
nent of the unit outward normal to the boundary at a boundary point x′. Skij(x′,x)
Chapter 5: A New Subregion Boundary Element Technique 92
and Dkij(x′,x) are linear combinations of derivatives of Tij(x
′,x) and Uij(x′,x), re-
spectively. The displacement integral equation (5.1) and the traction integral equa-
tion (5.2) are the governing equations to be solved for the displacement on the outer
boundary and the relative displacement on the crack faces.
We consider a three-subregion problem shown in Figure 5.1. In order to solve the
integral equations numerically, the boundary is discretised into a series of elements
on which displacement and traction components are written in terms of their values
at the nodal points. There are n1, n2 and n3 nodes placed on outer boundaries of the
subregions, m12 and m23 nodes on the interface between subregions, and nc nodes on
the crack face. Let ui and ti denote the nodal displacement and traction vectors on
Figure 5.1: A three subregion medium
boundary Γi respectively. Then, for the non-cracked subregions Ω1 and Ω3, the BEM
Chapter 5: A New Subregion Boundary Element Technique 93
equations can be written together in matrix form:(
h11 h12
h21 h22
)(
u1
u2
)
=
(
g11 g12
g21 g22
)(
t1
t2
)
(5.3)
and(
h77 h78
h87 h88
)(
u7
u8
)
=
(
g77 g78
g87 g88
)(
t7
t8
)
(5.4)
For the cracked subregion Ω2, referring to [66], the DBEM equations can be written
together in matrix form:
h33 h34 h35 h36
h43 h44 h45 h46
h53 h54 h55 h56
h63 h64 h65 h66
u3
u4
u5
∆u6
=
g33 g34 g35 g36
g43 g44 g45 g46
g53 g54 g55 g56
g63 g64 g65 g66
t3
t4
t5
t+6
(5.5)
where ∆u6 is the relative displacement vector on the crack surfaces and t+6 is the
traction vector on the upper crack surface.
The interface traction and displacement continuity conditions are
t2 = −t3, u2 = u3 (5.6)
between Ω1 and Ω2, and
t5 = −t7, u5 = u7 (5.7)
between Ω2 and Ω3.
In each subregion, we have more unknowns than the number of equations. To
solve the problem in the subregion level, we will have to use iterative methods which
are time consuming. A common practice would be to form a global matrix system
using the interface boundary conditions, such as in [18, 20, 59]. However, the resulting
matrix system is large that it is not cost effective. We propose a new subregion BEM
to form a smaller global matrix system. The differences between the current approach
and other approaches are discussed in the next section.
Chapter 5: A New Subregion Boundary Element Technique 94
5.3 Comparison of Subregion BEM Techniques
5.3.1 The Traditional Method
The traditional subregion BEM generate an N × N final matrix system based upon
equations (5.3), (5.4) and (5.5), and the interface conditions (5.6) and (5.7).
Au = Bt (5.8)
where
A =
h11 h12 0 0 0 0 −g12 0
h21 h22 0 0 0 0 −g22 0
0 h33 h34 h35 h36 0 g33 −g35
0 h43 h44 h45 h46 0 g43 −g45
0 h53 h54 h55 h56 0 g53 −g55
0 h63 h64 h65 h66 0 g63 −g65
0 0 0 h77 0 h78 0 g77
0 0 0 h87 0 h88 0 g87
, B =
g11 0 0 0
g21 0 0 0
0 g34 g36 0
0 g44 g46 0
0 g54 g56 0
0 g64 g66 0
0 0 0 g78
0 0 0 g88
u =(
u1 u2 u4 u5 ∆u6 u8 t2 t5
)T
, t =(
t1 t4 t+6 t8
)T
and N = 2(n1 + n2 + n3 + 2m12 + 2m23 + nc). The above matrix system may be
solved once the boundary conditions are prescribed. However, much computing time
and memory cost are required.
5.3.2 Kita & Kamiya’s Method
Both sides of equations (5.3), (5.4), and (5.5) are multiplied by the inverses of right
hand side matrices, respectively, to obtain the following boundary matrix equations(
a11 a12
a21 a22
)(
u1
u2
)
=
(
t1
t2
)
(5.9)
Chapter 5: A New Subregion Boundary Element Technique 95
(
a77 a78
a87 a88
)(
u7
u8
)
=
(
t7
t8
)
(5.10)
and
a33 a34 a35 a36
a43 a44 a45 a46
a53 a54 a55 a56
a63 a64 a65 a66
u3
u4
u5
∆u6
=
t3
t4
t5
t+6
(5.11)
where aij are the corresponding assembled matrices. Once the interface conditions
(5.6) and (5.7) are applied, Kita & Kamiya’s method then generates a N × N matrix
system
a11 a12 0 0 0 0
a21 a22 + a33 a34 a35 a36 0
0 a43 a44 a45 a46 0
0 a53 a54 a55 + a77 a56 a78
0 a63 a64 a65 a66 0
0 0 0 a87 0 a88
u1
u2
u4
u5
∆u6
u8
=
t1
0
t4
0
t+6
t8
(5.12)
where N = 2(n1 + n2 + n3 + m12 + m23 + nc). Once the interface displacement
components are solved from (5.12), the traction components can be calculated from
the systems of equations (5.9), (5.10) and (5.11). A smaller matrix system needs to
be solved compared with the traditional subregion BEM. However, three matrices of
the sizes N1 × N1, N2 × N2 and N3 × N3 need to inverted before the final matrix
can be established, where N1 = 2(n1 + m12), N2 = 2(n2 + m12 + m23 + nc) and
N3 = 2(n3 + m23). Since the number of operations required for inverting an n × n
matrix is of order n3, it offsets the savings provided by having a smaller global matrix.
That is the drawback of this method.
Chapter 5: A New Subregion Boundary Element Technique 96
5.3.3 The Proposed Method
In this section we propose a method that allows us to solve most of the problem at
the subregion level. In each subregion, we use the outer boundary conditions and
interface traction to represent the interface displacement. Upon applying the interfa-
cial equilibrium and compatibility conditions, we obtain a global matrix equation. In
this equation the interface traction is the only unknown. Once the matrix equation
is solved, the rest of the unknowns can then be solved at the subregion level on a
need to know basis. The interface traction matrix equation can be formed by a direct
inverse matrix method or a domain domain decomposition method.
The Direct Inverse Matrix Method
In the direct inverse matrix method [64], we multiply both sides of equations (5.3),
(5.4), and (5.5) by the inverse matrices of left hand side matrices to obtain
(
u1
u2
)
=
(
c11 c12
c21 c22
)(
t1
t2
)
(5.13)
(
u7
u8
)
=
(
c77 c78
c87 c88
)(
t7
t8
)
(5.14)
u3
u4
u5
∆u6
=
c33 c34 c35 c36
c43 c44 c45 c46
c53 c54 c55 c56
c63 c64 c65 c66
t3
t4
t5
t+6
(5.15)
where cij are the corresponding assembled matrices. Combing equations (5.13), (5.14)
and (5.15) with the equilibrium and compatibility conditions (5.6) and (5.7) on the
interface between the subregions, we obtain a 2(m12 + m23)× 2(m12 + m23) interface
Chapter 5: A New Subregion Boundary Element Technique 97
matrix equation
(
c22 + c33 c35
c53 c55 + c77
)(
t3
t5
)
=
(
c21 −c34 −c36 0
0 −c54 −c56 c78
)
t1
t4
t+6
t8
(5.16)
Once the interface traction components t3 and t5 are solved from (5.16), the displace-
ments can be calculated from equations (5.13), (5.14) and (5.15).
If part of the crack faces come to contact, instead of recalculating equation (5.5),
we only need to add an extra contact boundary condition on the final matrix equation
(5.16). Let ∆u6 = ∆uc6 + ∆uo
6, where ∆uc6 is the relative displacement vector of the
contact part on the crack faces and ∆uo6 is that of the open part on the crack faces.
Then the extra boundary condition is ∆uc6 = 0. For example, if the crack is fully
closed, the extra contact boundary condition can be obtained from (5.15), which is
0 = ∆u6 = ( c63 c64 c65 c66 )( t3 t4 t5 t+6 )T (5.17)
Therefore, the modified final matrix equation is
c22 + c33 c35 c36
c53 c55 + c77 c56
−c63 −c65 −c66
t3
t5
t+6
=
c21 −c34 0
0 −c54 c78
0 c64 0
t1
t4
t8
(5.18)
The Domain Decomposition Method
Instead of finding the inverse matrices directly, we can use the domain decomposition
method [51, 54, 83, 95] to form the interface traction matrix equation. In (5.3) and
(5.4), the outside boundary displacements can be expressed as
ui = h−1ii (giiti + gijtj − hijuj) (5.19)
Chapter 5: A New Subregion Boundary Element Technique 98
The displacements and tractions on the interface can then be easily written as
uj = Djjtj + Qj (5.20)
where
C−1jj = (hjj − hjih
−1ii hij)
−1
Djj = C−1jj (gjj − hjih
−1ii gij)
Qj = C−1jj (gji − hjih
−1ii gii)ti
and i = 1, j = 2 in Ω1, and i = 8, j = 7 in Ω3.
Rearrange (5.5) to allow terms (u4, ∆u6) together, these vectors can be expressed
as
(
u4
∆u6
)
= A−1
(
g43t3 + g44t4 + g46t+6 + g45t5 − h43u3 − h45u5
g63t3 + g64t4 + g66t+6 + g65t5 − h63u3 − h65u5
)
(5.21)
where
A−1 =
(
h44 h46
h64 h66
)−1
Let
B =
(
g44 g46
g64 g66
)
The unknown interface terms, i = 3 on the interface Ω1
⋂
Ω2 and i = 5 on the
interface Ω2
⋂
Ω3, can then be written as
Ci3u3 − Di3t3 + Ci5u5 − Di5t5 = Qi (5.22)
where
Ci3 = hi3 − (hi4,hi6)A−1
(
h43
h63
)
Chapter 5: A New Subregion Boundary Element Technique 99
Di3 = gi3 − (hi4,hi6)A−1
(
g43
g63
)
Ci5 = hi5 − (hi4,hi6)A−1
(
h45
h65
)
Di5 = gi5 − (hi4,hi6)A−1
(
g45
g65
)
Qi = [(gi4,gi6) − (hi4,hi6)A−1B]
(
t4
t+6
)
Upon applying the interface conditions (5.6) and (5.7) and substituting (5.20) into
(5.22), we obtain a 2(m12 + m23) × 2(m12 + m23) interface matrix equation,
(
C33D22 + D33, C35D77 + D35
C53D22 + D53, C55D77 + D55
)(
t2
t7
)
=
(
Q3 − C33Q2 − C35Q7
Q5 − C53Q2 − C55Q7
)
(5.23)
Once the interface tractions t2(= −t3) and t7(= −t5) are solved from (5.23), the
displacements can be calculated from the systems of equations (5.19), (5.20), and
(5.21).
The domain decomposition method has the advantage for cases where crack exten-
sion is modelled with new boundary elements. When boundary conditions changed
only part of the coefficient matrices need to be recalculated.
Both methods, the direct inverse matrix method and the domain decomposition
method, will derive the same final matrix equations. The direct inverse matrix method
is straightforward in formulation. It is useful for analyses when boundary conditions
remain the same. It is also useful when crack faces come to partial or full contact
because only the final matrix system needs to be modified. On the other hand,
in using the domain decomposition method, one does not need to solve as many
inverse matrices compared to the use of the direct inverse matrix method. It would
Chapter 5: A New Subregion Boundary Element Technique 100
provide further savings in computing time. Therefore, it is more efficient under general
boundary conditions.
The proposed method is suitable for parallel computation because the coefficient
matrix for each subregion can be calculated independently. This implementation
achieves a perfect scaling of the memory consumed per process, since each process
needs to access only a dedicated subregion of the entire domain. It enables us to
refine the boundary conditions in any subregion without having to recalculate others.
This feature is very useful for numerical analysis of crack extension, and cases where
crack faces come into contact.
Our method provides significant savings in computational time and memory usage.
Unlike other methods, which calculate the unknown components on the boundaries
and interfaces at the same time, only traction components on the interfaces are cal-
culated in the final matrix. As a result, the size of our final matrix is the number
of nodal degrees of freedom over the interfaces, and it is much smaller than those of
other methods.
Above all, our method has the advantage over the others because the reduction of
final matrix size and the ability of accessing each subregion independently. It results
in savings of the computing time and memory usage.
5.4 Numerical Results
In order to demonstrate the accuracy and efficiency of the proposed method and to
show its possible applications, several examples are presented here. All the calcula-
tions are carried out under plane strain conditions with a tensile loading T .
In the first example, consider the stress analysis of perfectly bonded dissimilar
Chapter 5: A New Subregion Boundary Element Technique 101
elastic semi-strips, as illustrated in Figure 5.2, where h1 = h2 = 2w. 30 elements
Figure 5.2: A perfectly bonded dissimilar elastic semi-strip
are placed on the outer boundary of each semi-stripe, and 20 elements on the inter-
face. Two cases are considered, with (E2/E1, ν1, ν2) = (9.0, 0.5, 0.5) and (3.0, 0.5, 0.5),
where E1 and E2 are the Young’s moduli, and ν1 and ν2 are the Poisson’s ratios, re-
spectively. The normalised normal stress distribution on the interface are shown in
Figure 5.3. The results agree well with published results [17].
Chapter 5: A New Subregion Boundary Element Technique 102
x/w
Nor
mal
ised
norm
alst
ress
dist
ribu
rtio
non
inte
rfac
e
0.5 0.6 0.7 0.8 0.9 10.9
1
1.1
1.2
1.3
1.4
1.5(a)(b)(c)(d)
Figure 5.3: Normalised normal stress distribution on the interface(E2/E1, ν1, ν2) = (9.0, 0.5, 0.5): (a) Ref. [17], (b) the present method(E2/E1, ν1, ν2) = (3.0, 0.5, 0.5): (c) Ref. [17], (d) the present method
Chapter 5: A New Subregion Boundary Element Technique 103
Consider, now, a three layered plate with a crack in the middle layer as shown
in Figure 5.4, where h1 = h2 = h3 = 0.5w, a/w = 0.1. The middle layer with
Figure 5.4: A three-layer plate with a centre crack
shear modulus µ2 and Poisson’s ration ν2 is perfectly bonded between two layers
having identical elastic properties µ1 = µ3 and ν1 = ν3. A crack of length 2a is
located l distance away from the top layer. Discontinuous quadratic elements are
used to discretise the boundaries, 36 elements on each subregion and 6 elements on
the crack surface. The stress intensity factor is normalised with respect to K0 =
T√
πa, calculated for various ratios of µ1/µ2 while ν1 = ν2 = ν3 = 0.3. Figure
5.5 shows the normalised mode I stress intensity factor versus the ratio of the shear
moduli (µ1/µ2) for various values of l/h2. The stress intensity factor increases as the
shear modulus ratio (µ1/µ2) decreases. It should be mentioned that the calculated
normalised stress intensity factor approaches 1 when the ratio 2a/h2 is small enough.
Chapter 5: A New Subregion Boundary Element Technique 104
It is expected, as this case is equivalent to an infinite homogeneous plate with a
central crack. Compared with the results calculated using other methods, the present
results match those of the traditional BEM method within eight decimal places.
µ1/µ2
Nor
mal
ised
mod
eI
SIF
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50.8
0.9
1
1.1
1.2
1.3
1.4
1.5
1.6(a)(b)(c)(d)(e)
Figure 5.5: Normalised mode I stress intensity factor (SIF) on the three-layer platewith a centre crack: (a) l/h2 = 0.1, (b) l/h2 = 0.2, (c) l/h2 = 0.3, (d) l/h2 = 0.4, (e)l/h2 = 0.5
Chapter 5: A New Subregion Boundary Element Technique 105
Finally, consider a three layered plate with two identical co-linear cracks of length
2a located symmetrically in the middle layer as shown in Figure 5.6, where h1 =
h2 = h3 = 0.5w, a/w = 0.1. The distance between the centres of the cracks is
2d = 2.4a. The elastic properties are taken as µ1 = µ3 and ν1 = ν2 = ν3 = 0.3.
36 discontinuous boundary elements are placed on each subregion boundary, and 6
Figure 5.6: A three-layer plate with two identical co-linear cracks
discontinuous elements on each crack. Again the normalised mode I stress intensity
factor is calculated for various ratios of the shear moduli (µ1/µ2), the results are
shown in Figure 5.7 for tip A and Figure 5.8 for tip B. It is noted that the stress
intensity factor increases as the shear modulus ratio (µ1/µ2) decreases. Further, due
to the interaction between the two cracks, the mode I stress intensity factor at the
crack tip A is always smaller than that at the crack tip B.
Chapter 5: A New Subregion Boundary Element Technique 106
µ1/µ2
Nor
mal
ised
mod
eI
SIF
attip
A
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 51
1.1
1.2
1.3
1.4
1.5
1.6
1.7(a)(b)(c)(d)(e)
Figure 5.7: Normalised mode I SIF at tip A on the three-layer plate with two co-linearcracks: (a) l/h2 = 0.1, (b) l/h2 = 0.2, (c) l/h2 = 0.3, (d) l/h2 = 0.4, (e) l/h2 = 0.5
Chapter 5: A New Subregion Boundary Element Technique 107
µ1/µ2
Nor
mal
ised
mod
eI
SIF
attip
B
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 51
1.1
1.2
1.3
1.4
1.5
1.6
1.7(a)(b)(c)(d)(e)
Figure 5.8: Normalised mode I SIF at tip B on the three-layer plate with two co-linearcracks: (a) l/h2 = 0.1, (b) l/h2 = 0.2, (c) l/h2 = 0.3, (d) l/h2 = 0.4, (e) l/h2 = 0.5
Chapter 5: A New Subregion Boundary Element Technique 108
5.5 Conclusion
Three examples were examined using the new subregion boundary element technique.
Compared with other methods, our method was shown to be very effective and ac-
curate for the boundary element analysis of an object composed of subregions. Our
method achieved a substantial reduction in the number of numerical computations.
For example, for a problem involving three subregions, only a small number of ele-
ments are used on the boundary of each subregion, the number of computations is
reduced by 43%. This shows that our method has the advantages of significantly
reducing computing time and memory usage. Furthermore, it would be advantageous
to use the method for solving problems with changing boundary conditions such as
those encountered in crack growth analysis. Because each calculation in each sub-
region is independent, only those equations related to the regions where changes of
crack boundary conditions occur need to be recalculated.
Chapter 6
A Subregion DRBEM Formulation
for the Dynamic Analysis of Two
Dimensional Cracks
The dual reciprocity boundary element method employing the step by step time inte-
gration technique is developed to analyse two-dimensional dynamic crack problems.
In this method the equation of motion is expressed in boundary integral form us-
ing elastostatic fundamental solutions. In order to transform the domain integral
into an equivalent boundary integral, a general radial basis function is used for the
derivation of the particular solutions. The dual reciprocity boundary element method
is combined with an efficient subregion boundary element method to overcome the
difficulty of a singular system of algebraic equations in crack problems. Dynamic
stress intensity factors are calculated using the discontinuous quarter point elements.
Several examples are presented to show the formulation details and to demonstrate
the computational efficiency of the method.
109
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 110
6.1 Introduction
Stresses and displacements in dynamic loading can be greatly different to those as-
sociated with static loading. The equation of motion is also more complex than the
corresponding equilibrium equation for the same configuration and material class.
Therefore, structures with arbitrary shapes and time dependent boundary conditions
need to be analysed using numerical methods. One numerical method which has
had great success in solving elastodynamic problems is the boundary element method
(BEM) [34]. Various numerical techniques using BEM are usually implemented by
using either the time domain method, the Laplace transform method or the dual
reciprocity boundary element method (DRBEM).
The time domain method combined with the dual boundary element method was
used by Fedelinski et al. [42, 45]. A hypersingular formulation for time domain anti-
plane elastodynamic problems was developed by Gallego and Dominguez [46]. And a
three dimensional time domain boundary integral equation method was presented by
Zhang and Gross [104] for transient elastodynamic crack analysis. All of these cur-
rent methods possess advantages and disadvantages. For example, the time domain
method requires time and space integrations. These integrations have a complicated
form, resulting in the time domain method being difficult to implement. Furthermore,
this method is computationally intensive, as it requires a large amount of computation
time because of its double multiplication.
The Laplace transform method was used by Sladek and Sladek [88, 89] to solve
elastodynamic problems. The boundary value problems were formulated by the
boundary integral element method in the Laplace transform domain. This method
was combined with the numerical Laplace inversion algorithm of Durbin [37]. The
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 111
Laplace transform method combined with the dual boundary element method was
used by Fedelinski et al. [44] to solve two-dimensional dynamic crack problems. The
dynamic problems can then be solved with a single region formulation in the Laplace
transform domain. Three-dimensional dynamic problems using a similar method were
also solved by Wen et al. [98]. In the Laplace transform method, only the space in-
tegration is required, although this transform depends on complex Bessel functions.
The generation of equations in this method is time consuming because of the com-
plexity of the transformed fundamental solutions. The accuracy and efficiency of the
Laplace transform method depend on the level of difficulty of the numerical inversion
of the Laplace transform.
A considerable improvement on the time domain and Laplace transform methods
is the dual reciprocity boundary element method (DRBEM). In DRBEM, the equa-
tion of motion is expressed as a boundary integral equation by using the fundamental
solutions of elastostatics and the approximation of the acceleration. The fundamental
solutions of elastostatics are time independent, thus only space integration is required.
The acceleration can be obtained by approximation functions multiplied by unknown
time dependent coefficients. The approximation functions do not require integration.
Radial basis functions are chosen for the approximation functions and used for deter-
mining particular solutions of the displacement and traction vectors in DRBEM. This
method has an easier formulation with direct application to dynamic problems. The
DRBEM combined with the dual boundary element method was used by Fedelinski et
al. [40, 41]. Various radial basis functions were implemented by Agnantiris et al. in
the solutions of the two dimensional and three dimensional symmetric elastodynamic
problems using the DRBEM [3, 4]. The vibration problem of a thin plate was solved
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 112
by Davies and Moslehy [33] using the DRBEM. However, the coincidence nodes on
the crack surfaces give rise to a singular system of algebraic equations. The unknown
time dependent coefficients cannot be calculated by the direct solution of this sys-
tem. This difficulty is usually overcome by using either the subregion BEM or the
dual boundary element method, see Fedelinski et al. [40].
Although DRBEM possesses certain obvious advantages over the other methods,
the algebraic singularities present in the method have not yet been adequately dealt
with. An efficient subregion BEM is generally needed to overcome the singular system
of algebraic equations on the coincidence crack nodes. In this chapter, we propose a
DRBEM formulation based on an advanced subregion boundary element technique
[64, 67] to address the singularity problem. A structure is divided into sub-domains
along crack surfaces. The interface equations are assembled using the interface equi-
librium and compatibility conditions. The treatment of time dependent functions
with the DRBEM is dealt directly in the time domain. The Houbolt method [78, 12]
will be used for the direct step by step time integration procedures. Since the system
of equations for each subregion are pre-solved independently, only the time depen-
dent equations need to be recalculated. This formulation is very effective due to the
reduction in the size of final matrix and in the number of intermediate steps. Numer-
ical examples are provided to demonstrate the accuracy and efficiency of the present
formulation. The effect of internal collocation points on numerical accuracy is also
examined.
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 113
6.2 Dual Reciprocity Boundary Element Method
Consider an elastic isotropic homogeneous body occupying any two dimensional do-
main Ω, enclosed by a boundary Γ. The differential equation of motion can be
expressed as follows:
σjk,k + bj − ρuj = 0 (6.1)
where σjk is the cartesian stress tensor, uj is the displacement vector, bj is the body
force, ρ denotes the mass density, subscript k proceeded by a comma denotes differen-
tiation with respect to the coordinate xk and dots denote differentiation with respect
to time. The equation can also be expressed in terms of the displacement
µuj,kk +µ
1 − 2νuk,kj + bj − ρuj = 0 (6.2)
where µ is the shear modulus and ν is the Poisson’s ratio.
In the absence of body forces, the boundary element formulation for elastodynamic
problems can be expressed as
ui(X′, τ) −
∫
Γ
Uij(X′,x)tj(x, τ)dΓ(x) +
∫
Γ
Tij(X′,x)uj(x, τ)dΓ(x)
= −ρ
∫
Ω
Uij(X′,X)uj(X, τ)dΩ(X) (6.3)
where ui is the displacement component in the i-direction at internal point X′ and
time τ , Uij(X′,x) and Tij(X
′,x) represent the static Kelvin displacement and traction
fundamental solutions, respectively. For a boundary point (i.e. X′ → x′), equation
(6.3) can be written as
cij(x′)uj(x
′, τ) −∫
Γ
Uij(x′,x)tj(x, τ)dΓ(x) +
∫
Γ
− Tij(x′,x)uj(x, τ)dΓ(x)
= −ρ
∫
Ω
Uij(x′,X)uj(X, τ)dΩ(X) (6.4)
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 114
where∫
− stands for a Cauchy principal value integral. Coefficient cij(x′) depends on
boundary point x′, on a smooth boundary cij(x′) = 1
2δij. The DRBEM will be used
to transform the domain integral in equation (6.4) into a boundary integral.
The acceleration uj at a point X and time τ is approximated as a sum of m
coordinate functions fn(X′′,X) multiplied by unknown time-dependent coefficients
αnj (τ):
uj(X, τ) =m∑
n=1
αnj (τ)fn(X′′,X) (6.5)
Equation (6.5) is valid over the whole domain. X′′ can be a boundary or a domain
point (see Figure 6.1). Using (6.5), the domain term of the displacement integral
Figure 6.1: Boundary and internal nodes
equation (6.4) can be written as
ρ
∫
Ω
Uij(x′,X)uj(X, τ)dΩ(X) = ραn
l (τ)
∫
Ω
Uij(x′,X)δljf
n(X′′,X)dΩ(X) (6.6)
where αnj (τ) = αn
l (τ)δlj, and δlj is the Kronecker delta. In order to transform the
domain integral into equivalent boundary integral, we need to find particular solutions
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 115
unlj and tnlj of the following differential equation:
σljk,k − δljfn = 0, (6.7)
in the absence of body forces. The domain integral in equation (6.6) is rewritten as
ραnl (τ)
∫
Ω
Uij(x′,X)δljf
n(X′′,X)dΩ(X)
= −ραnl (τ)
[
cij(x′)un
lj(X′′,x′) −
∫
Γ
Uij(x′,x)tnlj(X
′′,x)dΓ(x)
+
∫
Γ
− Tij(x′,x)un
lj(X′′,x)dΓ(x)
]
(6.8)
where the particular displacement unlj(X
′′,x) and traction tnlj(X′′,x) are given in Sec-
tion 6.3. Substituting equation (6.8) into equation (6.4), the displacement boundary
equation can be written as
cij(x′)uj(x
′, τ) −∫
Γ
Uij(x′,x)tj(x, τ)dΓ(x) +
∫
Γ
− Tij(x′,x)uj(x, τ)dΓ(x)
= ραnl (τ)
[
cij(x′)un
lj(X′′,x′) −
∫
Γ
Uij(x′,x)tnlj(X
′′,x)dΓ(x)
+
∫
Γ
− Tij(x′,x)un
lj(X′′,x)dΓ(x)
]
(6.9)
In order to obtain a solution of the elastodynamic problem, the boundary is divided
into small boundary elements. We obtain a set of equations written in matrix form
as
Hu − Gt = ρ(
Hu − Gt)
α (6.10)
The relationship between u and α is established by applying equation (6.5) to every
boundary and domain node. The resulting set of equations can be written in matrix
form:
u(x, τ) = F(X′′,x)α(τ) (6.11)
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 116
where the elements of the matrix F(X′′,x) are the values of function fn(X′′,x) at all
m nodes. The unknown coefficients α can be expressed as
α(τ) = F−1(X′′,x)u(x, τ) (6.12)
If the coincident nodes of the crack are used in the approximation of the acceleration
field, the system of equations (6.11) will be singular. Therefore, the coefficients α
cannot be calculated by the direct solution of this system. A new subregion BEM
[67] is used to overcome this difficulty (see Chapter 5). Substituting equation (6.12)
into equation (6.10), we obtain
Hu − Gt = ρ(
Hu − Gt)
F−1u (6.13)
Equation (6.13) can be written in the compact form as
Mu + Hu = Gt (6.14)
where
M = −ρ(
Hu − Gt)
F−1 (6.15)
The system of equations (6.14) is modified, according to the boundary conditions,
and can be solved using a direct integration method.
The Houbolt integration scheme [12, 78] is used to approximate the acceleration
components in terms of the displacement components. The time span under consid-
eration, T , is divided into N equal time intervals ∆τ (T = N · ∆τ). Assuming that
the solution of (6.14) is known at 0, ∆τ, 2∆τ, . . . , T , the following algorithm is used
to obtain the acceleration at time τ +∆τ , and is derived by using the finite difference
expression
uτ+∆τ =1
(∆τ)2(2uτ+∆τ − 5uτ + 4uτ−∆τ − uτ−2∆τ ) (6.16)
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 117
The general system of equation (6.14) at time τ + ∆τ can be written as
Muτ+∆τ + Huτ+∆τ = Gtτ+∆τ (6.17)
Upon substituting (6.16) into (6.17) one obtains
[
2
(∆τ)2M + H
]
uτ+∆τ = Gtτ+∆τ + M1
(∆τ)2(5uτ − 4uτ−∆τ + uτ−2∆τ ) (6.18)
or
Auτ+∆τ = Gtτ+∆τ + M (a1uτ + a2uτ−∆τ + a3uτ−2∆τ ) (6.19)
where A = 2/(∆τ)2M + H, a1 = 5/(∆τ)2, a2 = −4/(∆τ)2, and a3 = 1/(∆τ)2.
The resulting boundary integral equations on a subregion at time τ+∆τ , rewritten
from equation (6.19), are
A+u+τ+∆τ = G+t+
τ+∆τ + M+(
a1u+τ + a2u
+τ−∆τ + a3u
+τ−2∆τ
)
(6.20)
on the subregion with the upper crack face, Ω+ and
A−u−τ+∆τ = G−t−τ+∆τ + M−
(
a1u−τ + a2u
−τ−∆τ + a3u
−τ−2∆τ
)
(6.21)
on the subregion with the lower crack face, Ω− (see Figure 6.2). Using the new
subregion boundary element method in Chapter 5, we combine equations (6.20) and
(6.21) with the interface equilibrium and compatibility conditions
t+2 = −t−1 , t+
4 = −t−3 and u+2 = u−
1 , u+4 = u−
3 (6.22)
to obtain the following interface traction matrix
[
b21 b22
b41 b42
][
t+2
t+4
]
τ+∆τ
=
[
b23 b24 b25 b26
b43 b44 b45 b46
]
t+1
t+3
t−2
t−4
τ+∆τ
+
[
w1
w2
]
τ+∆τ
(6.23)
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 118
Figure 6.2: Crack on the body for subregion method
where bij, and wi are matrix coefficients from A+,−, G+,−, M+,−, and ai. Once
the unknown interface traction components are solved from equation (6.23), the dis-
placement components can be obtained from equations (6.20) and (6.21) subject to
boundary conditions.
6.3 Derivation of Particular Solutions
In this section, a general method for analytically obtaining a particular solution of
equation (6.7) for two dimensional problems is described as in [3, 4]. A polynomial
radial basis function is used for deriving the solutions of the particular displacement
and traction vectors. The expression for the approximating radial basis function
fn(X′′,x) is
fn(X′′,x) =
q∑
p=0
Ap [r(X′′,x)]p
(6.24)
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 119
where r(X′′,x) is the distance between the point X′′ and the boundary point x.
The point X′′ can be located in the domain or at the boundary. The particular
displacement solution unlj(X
′′,x) is obtained from equation (6.7) using equation (6.24)
unlj(X
′′,x) =1
µ
q∑
p=0
Ap
[
(4p + 14) − 4(p + 4)ν
4(1 − ν)(p + 2)2(p + 4)rp+2δlj
− 1
2(1 − ν)(p + 2)(p + 4)rp+2r,lr,j
]
(6.25)
The corresponding expression for the particular traction solution tnlj(X′′,x) is
tnlj(X′′,x) =
q∑
p=0
Ap
[
1
2(1 − ν)
(
1 − 2ν
p + 2+
1
p + 4
)
rp+1 (r,lnj + r,knkδlj)
− prp+1
(1 − ν)(p + 2)(p + 4)r,knkr,lr,j
+1
2(1 − ν)
(
−1 − 2ν
p + 2+
1
p + 4
)
rp+1nlr,j
]
(6.26)
6.4 The Dynamic Stress Intensity Factors
One of the important parameters in dynamic fracture mechanics analysis is the dy-
namic stress intensity factor (DSIF) since it characterizes the stress field in the vicinity
of the crack and controls crack growth. In the present approach it is calculated di-
rectly from the displacement components near the crack tip. In order to improve the
accuracy in calculating the dynamic stress intensity factors, the discontinuous quarter
point element method is applied [16, 71]. The dynamic stress intensity factors have
the same definition as in a static problem except that they are time dependent. The
dynamic stress intensity factors for mode I and mode II can be calculated from the
crack opening displacement as follows
KI,II =µ
κ + 1
√
2π
r∆un,t(r, τ) (6.27)
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 120
where κ = 3−4ν for plane strain, r is the distance from crack tip to the nearest node
on the crack face. ∆un(r, τ) and ∆ut(r, τ) denote the relative normal and tangential
displacement at a distance r to the crack tip at time τ respectively.
6.5 Numerical Examples
In order to demonstrate the accuracy and efficiency of the present method, we present
three examples. The first example shows an application to a rectangular plate with
a central crack. The second example considers the analysis of a rectangular plate
with a central slant crack. The solutions of these two examples are compared with
other published solutions [7, 8, 43, 76]. The third example shows the analysis of a
rectangular plate with an internal kinked crack. The dynamic stress intensity factors
are obtained from crack opening displacement calculations by using the discontinuous
quarter point element method. In each case the structure is subjected to a uniform
tension
σ(τ) = σ0H(τ) (6.28)
where H(τ) is the Heaviside step function. The plate is under plane strain condition.
The dynamic stress intensity factors are normalised with respect to
K0 = σ0
√πa (6.29)
where a defines the half cracklength, except in the third example, where a different
normalising factor is more suitable.
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 121
6.5.1 A Rectangular Plate with a Central Crack
Consider a rectangular plate of length 2h = 40 mm and width 2w = 20 mm containing
a central crack. The crack has length 2a = 4.8 mm, as shown in Figure 6.3. The
Figure 6.3: Rectangular plate with a central crack
material properties are: the shear modulus µ = 7.692 × 1010Pa, Poisson’s ratio
ν = 0.3, and density ρ = 5 × 103kg · m−3. Discontinuous quadratic elements are
used to discritise the boundary. There are 30 elements and 21 additional domain
points used on each subregion, in which 6 elements are on the crack face. The time
step ∆τ = 0.2 µs is used here. f(X′′,x) = 1 + r(X′′,x) is used as the approximation
function of the particular solutions. The mode I dynamic stress intensity factor vs.
time τ is plotted in Figure 6.4. It is seen that at time τ = 6.8 µs, the first peak
value of the normalised mode I dynamic stress intensity factor is 2.52, and compares
well with the results published in Reference [6], where the peak values are 2.54 at
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 122 Time 6.6 µs for the dual reciprocity method and 2.67 at time 6.4 µs for the Laplace transform
method, respectively.
Figure 3.7: Normalised mode I DIF for the rectangular plate with a central crack (a) the present method, (b) the dual reciprocity method [6] and (c) the Laplace transform method [6]
Please see print copy for Figure 6.4
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 123
6.5.2 A Rectangular Plate with a Central Slant Crack
Consider a rectangular plate of length 2h = 60 mm and width 2w = 30 mm containing
a central slant crack. The crack has length 2a = 14.14 mm slanted with an angle
θ = 45 from the horizontal direction, as shown in Figure 6.5. The material properties
Figure 6.5: Rectangular plate with a central slant crack
are: the shear modulus µ = 7.692 × 1010Pa, Poisson’s ratio ν = 0.3, and density
ρ = 5× 103kg ·m−3. The discontinuous quadratic elements are used to discretise the
boundary. There are 36 elements and 25 additional domain points on each subregion,
in which 6 elements are on the crack face. The time step ∆τ = 0.3 µs is used
here. f(X′′,x) = 1 + r(X′′,x) is used as the approximation function of the particular
solutions. The mode I dynamic stress intensity factor vs. time τ is plotted in Figure
6.6. The first peak value of the normalised mode I dynamic stress intensity factor
is 1.26 at τ = 10.8 µs, and compares well with the dual reciprocity and the Laplace
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 124
transform methods whose peak values are 1.27 at time 11.6 µs and 1.31 at time 11.3 µs,
respectively. The mode II dynamic stress intensity factor vs. time τ is plotted in Figure 6.7. At τ =
10.2 µs, the first peak value of the normalized mode II dynamic stress intensity factor is 1.50, and
compares well with the results published in Reference [6], where the peak values are 1.44 at time
10.3 µs for the dual reciprocity method and 1.51 at time 10.2 µs for the Laplace transform method,
respectively.
Figure 6.6: Normalised mode I DSIF for the rectangular plate with a central slant crack (a) the present method, (b) the dual reciprocity method [6] and (c) the Laplace transform method [6]
Please see print copy for Figure 6.6
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 125
Figure 6.7: Normalised mode II DSIF for the rectangular plate with a central slant crack (a) the present method, (b) the dual reciprocity method [6] and (c) the Laplace transform method [6]
Please see print copy for Figure 6.7
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 126
6.5.3 A Rectangular Plate with an Internal Kinked Crack
Consider a rectangular plate of length 2h = 40 mm and width 2w = 20 mm containing
an internal kinked crack, shown in Figure 6.8. One of the segments of the crack is
Figure 6.8: Rectangular plate with an internal kinked crack
horizontal with length a while the other segment makes an angle of 45 with the
horizontal and has a length b. The horizontal projection of the total crack is given
by 2c = a +√
2b/2. The kink of the crack is at the centre of the plate. The material
properties are: the shear modulus µ = 7.692 × 1010Pa, Poisson’s ratio ν = 0.3, and
density ρ = 5 × 103kg · m−3. The time step ∆τ = 0.3 µs is used here. f(X′′,x) =
1 + r(X′′,x) is used as the approximation function of the particular solutions. Three
cases were considered, b/a = 0.2, 0.4, and 0.6, with a/w = 0.1. The dynamic stress
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 127
intensity factors are normalised with respect to
K0 = σ0
√πc (6.30)
They were obtained for both tips A and B, with a boundary element mesh of 45
discontinuous quadratic elements and 19 additional domain points on each subregion,
in which the horizontal and the inclined segments of the crack were discretised with 5
and 4 discontinuous quadratic elements on each crack face, respectively. The results
are presented in Figure 6.9 for tip A. The first peak value of the normalised mode
I dynamic stress intensity factor is 2.48, 2.47, and 2.47 for b/a = 0.2, 0.4, and 0.6,
respectively, at τ = 6.6 µs. The results for tip B are presented in Figure 6.10 and
6.11. The first peak values of the normalised mode I dynamic stress intensity factor
are 1.53 at τ = 6.6 µs for b/a = 0.2, 1.32 at τ = 6.9 µs for b/a = 0.4 and 1.26 at
τ = 6.9 µs for b/a = 0.6; the peak value decreases when the ratio b/a increases. The
first peak value of the normalised mode II dynamic stress intensity factor is −1.74,
−1.81, and −1.85 for b/a = 0.2, 0.4, and 0.6, respectively, at τ = 6.3 µs; the peak
value increases when the ratio b/a increases.
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 128
τ(µs)
Nor
mal
ised
mod
eI
DS
IFat
tipA
0 2 4 6 8 10 12 14-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
2.2
2.4
2.6(a)(b)(c)
Figure 6.9: Normalised mode I DSIF for the rectangular plate with an internal kinkedcrack at tip A (a) b/a = 0.2, (b) b/a = 0.4, and (c) b/a = 0.6
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 129
τ(µs)
Nor
mal
ised
mod
eI
DS
IFat
tipB
0 2 4 6 8 10 12 14-0.2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8(a)(b)(c)
Figure 6.10: Normalised mode I DSIF for the rectangular plate with an internalkinked crack at tip B (a) b/a = 0.2, (b) b/a = 0.4, and (c) b/a = 0.6
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 130
τ(µs)
Nor
mal
ised
mod
eII
DS
IFat
tipB
0 2 4 6 8 10 12 14-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2(a)(b)(c)
Figure 6.11: Normalised mode II DSIF for the rectangular plate with an internalkinked crack at tip B (a) b/a = 0.2, (b) b/a = 0.4, and (c) b/a = 0.6
Chapter 6: A Subregion DRBEM Formulation for Dynamic Analysis 131
6.6 Conclusion
In this chapter the dual reciprocity boundary element method employing the step
by step time integration technique is developed to analyse two-dimensional dynamic
crack problems. The dual reciprocity boundary element method was used to trans-
form the domain integrals into equivalent boundary integrals in the two dimensional
boundary element analysis. An efficient subregion technique is used to overcome the
difficulty of a singular system of algebraic equations in crack problems. The system
of ordinary differential equations in time is solved using the step by step finite differ-
ence method. The proposed method is more efficient than traditional BEM because
it significantly reduces the size of the final matrix system. Further, as the system
of equations for each subregion is pre-solved independently, only the time dependent
equations need to be recalculated, which is advantageous when using the step by step
time integration approach. The dynamic stress intensity factors are calculated from
the crack tip opening displacement by using the discontinuous quarter point method.
The calculated results for the examples compared well with those of published solu-
tions. It is, therefore, shown that the present formulation can be effectively applied
to study elastodynamic fracture problems.
Chapter 7
Conclusion
In this thesis, we studied the advanced boundary element method for fracture mechan-
ics, including static and dynamic problems. Static problems were solved by using the
dual boundary element method, which based on the works in [66]. Normally, the stress
intensity factor is obtained by using discontinuous quarter point element method. In
this work, we presented a special crack tip element method, which provides similar
accuracy as that of quarter point element method, but a much easier discritisation of
the crack face for evaluating stress intensity factors.
Further, a new subregion boundary element method was presented to solve com-
posite material problems, which based on the work in [64]. Similar composite prob-
lems are also solved by using domain decomposition method, which based on the
work in [67]. Dynamic fracture mechanics problems were solved by using the dual
reciprocity boundary element method, which based on the work in [65].
In chapters 2 and 3, we provided a detailed study of two dimensional linear elas-
tic crack problems under in-plane tensile and anti-plane shear loadings, respectively.
The difficulty in using the boundary element method directly is that the upper crack
132
Chapter 7: Conclusion 133
surface and lower crack surface are both located in the same mathematical posi-
tion. However, they have different physical boundary conditions. The dual boundary
element method was used to overcome this difficulty. In order to obtain accurate
results, the dual boundary element method was combined with the discontinuous
quarter point element method to correctly model the characteristics of displacements
near the crack tip. The improved efficiency of this method was due to the fact that the
crack problems could be solved with a single region formulation. The accuracy of the
method was also improved compared with published results. The error percentages
of the present numerical results were less than 1 per cent.
In Chapter 4, we concerned with a special crack tip element method to evalu-
ate stress intensity factors. The calculated results were compared with those of the
discontinuous quarter point element method. Using the discontinuous quarter point
element method we need to choose the locations of the collocation points which de-
pend on the shape functions at the crack tip element. However, the special crack tip
element method does not require shifting of the collocation points at the crack tip
element. Therefore, the presented method is a more straight-forward technique than
discontinuous quarter point element method. Furthermore, both methods produced
results with similar accuracy.
In Chapter 5, we presented a new subregion boundary element method. Existing
methods generate a large system of equations, the solution process is time consuming
and computationally intensive. In the new method, only a relatively smaller matrix
system, the interface traction matrix system, needs to be solved. Therefore, a much
better computational efficiency is achieved. The embedded crack problem was inves-
tigated by combining the dual boundary element method with this new technique.
Chapter 7: Conclusion 134
In Chapter 6, dynamic fracture problems were studied using the dual reciprocity
boundary element method (DRBEM). The DRBEM is used to transform the domain
integral term, and polynomial radial basis functions provided the particular solutions
of both displacement and traction kernels. The study also employed the subregion
boundary element method, which was presented in Chapter 5. The time-dependent
differential equation was solved using a direct step-by-step procedure, and the inter-
face traction matrix system were solved by using an iterative procedure that updates
the estimate at each time step.
The work in this thesis could be extended to three-dimensional crack analysis.
However, due to the complexity of the three-dimensional geometry, it requires more
effective modelling strategy to achieve better computational efficiency. We could
combine the boundary element method (BEM) and the finite element method (FEM)
in future study. FEM models are very good at representing global behaviours, but
difficult to model small details and features such as cracks, whereas BEM can capture
local behaviours and small geometry details. By combining the techniques we can
reduce the computation time involved in crack modelling and obtain more accurate
predictions.
The idea of combining the BEM and FEM is to decompose the domain of interest
into sub-domains and to use the most appropriate technique for each sub-domain.
The key point in doing so is the coupling of the systems of equations on the inter
domain boundaries [1, 2, 14, 47]. The equations produced by the BEM and FEM
are expressed in terms of different variables and cannot be linked as they stand [38].
Conventional coupling methods may destroy the desirable features of the BEM and
FEM. Hence future work should include finding a suitable coupling method.
Bibliography
[1] State of the art in BEM/FEM coupling, International Journal of Boundary
Element Methods Communications 4 (1993), 58–67.
[2] State of the art in BEM/FEM coupling, International Journal of Boundary
Element Methods Communications 4 (1993), 94–104.
[3] J.P. Agnantiaris, D. Polyzos, and D.E. Beskos, Some studies on dual reciprocity
BEM for elastodynamic analysis, Computational Mechanics 17 (1996), 270–277.
[4] , Three-dimensional structural vibration analysis by the dual reciprocity
BEM, Computational Mechanics 21 (1998), 372–381.
[5] M. Ali and N.J. Altiero, Solution of the problem of a crack in a finite plane
region using an indirect boundary-integral method, Engineering Fracture Me-
chanics 11 (1979), 831–837.
[6] M.H. Aliabadi, Dynamic fracture mechanics, Computational mechanics publi-
cations, Southampton, 1995.
[7] M.H. Aliabadi and C.A. Brebbia, Advanced formulations in boundary element
methods, Computational mechanics publications, Southampton, 1993.
[8] , Advances in boundary element methods for fracture mechanics, Com-
putational mechanics publications, Southampton, 1993.
135
BIBLIOGRAPHY 136
[9] M.H. Aliabadi and D.P. Rooke, Numerical fracture mechanics, Computational
mechanics publications, Southampton, 1991.
[10] M.H. Aliabadi and A.L. Saleh, Fracture mechanics analysis of cracking in plain
and reinforced concrete using the boundary element method, Engineering Frac-
ture Mechanics 69 (2002), 267–280.
[11] T.L. Anderson, Fracture mechanics fundamentals and applications, CRC Press,
Boca Raton, 1995.
[12] K.J. Bathe, Finite element procedures, Prentice Hall, New Jersey, 1996.
[13] V.T. Bechel and A.K. Kaw, Fracture mechanics of composites with nonhomo-
geneous interphases and nondilute fiber volume fractions, International Journal
of Solids and Structures 31 (1994), 2053–2070.
[14] R.A. Bialecki, Z. Ostrowski, A.J. Kassab, Q. Yin, and E. Sciubba, Coupling
BEM, FEM and analytic solutions in steady-state potential problems, Engineer-
ing Analysis with Boundary Elements 26 (2002), 597–611.
[15] W.D. Biggs, The brittle fracture of steel, MacDonald and Evans, London, 1960.
[16] G.E. Blandford, A.R. Ingraffea, and J.A. Liggett, Two-dimensional stress in-
tensity factor computations using the boundary element method, International
Journal for Numerical Methods in Engineering 17 (1981), 387–404.
[17] D.B. Bogy, The plane solution for joined dissimilar elastic semistrips under
tension, Journal of Applied Mechanics (1975), 93–98.
[18] C.A. Brebbia and J. Dominguez, Boundary Elements: An Introductory Course,
Computational Mechanics Publications, Southampton and Boston, 1989.
[19] C.A. Brebbia and P. Georgiou, Combination of boundary and finite elements in
elastostatics, Applied Mathematical Modelling 3 (1979), 212–220.
BIBLIOGRAPHY 137
[20] C.A. Brebbia, J.C.F. Telles, and L.C. Wrobel, Boundary Element Techniques:
Theory and Applications in Engineering, Springer-Verlag, Berlin and New York,
1984.
[21] K.B. Broberg, Cracks and fracture, Academic Press, San Diego, 1999.
[22] D. Broek, Elementary engineering fracture mechanics, Martinus Nijhoff Pub-
lishers, Taipei, 1986.
[23] F.M. Burdekin and D.E.W. Stone, The crack opening displacement approach to
fracture mechanics in yielding materials, Journal of Strain Analysis 1 (1966),
145–153.
[24] W.H. Chen and C.S. Chang, Analysis of two-dimensional mixed-mode crack
problems by finite element alternating method, Computers and Structures 33
(1989), 1451–1458.
[25] W.H. Chen and T.C. Chen, An efficient dual boundary element technique for
a two-dimensional fracture problem with multiple cracks, International Journal
for Numerical Methods in Engineering 38 (1995), 1739–1756.
[26] C.R. Chiang, Slightly curved cracks in antiplane strain, International Journal
of Fracture 32 (1987), R63–R66.
[27] A.P. Cisilino and M.H. Aliabadi, Three-dimensional BEM analysis for fatigue
crack growth in welded components, International Journal of Pressure Vessels
and Piping 70 (1997), 135–144.
[28] , Three-dimensional boundary element analysis of fatigue crack growth
in linear and non-linear fracture problems, Engineering Fracture Mechanics 63
(1999), 713–733.
BIBLIOGRAPHY 138
[29] A.P. Cisilino, M.H. Aliabadi, and J.L. Otegui, A three-dimensional boundary el-
ement formulation for the elastoplastic analysis of cracked bodies, International
Journal for Numerical Methods in Engineering 42 (1998), 237–256.
[30] B. Cotterell and J.R. Rice, Slightly curved or kinked cracks, International Jour-
nal of Fracture 16 (1980), 155–169.
[31] S.L. Crouch, Solution of plane elasticity problems by the displacement discon-
tinuity method. I. Infinite body solution, International Journal for Numerical
Methods in Engineering 10 (1976), 301–343.
[32] T.A. Cruse, Boundary elements analysis in computational fracture mechanics,
Kluwer academic publishers, Dordrecht, 1988.
[33] T.W. Davies and F.A. Moslehy, Modal analysis of plates using the dual reci-
procity boundary element method, Engineering Analysis with Boundary Ele-
ments 14 (1994), 357–362.
[34] J. Dominguez, Boundary elements in dynamics, Computational Mechanics Pub-
lications, Southampton, 1993.
[35] J.J. Duga, W.H. Fisher, R.W. Buxbaum, A.R. Rosenfield, A.R. Burh, E.J.
Honton, and S.C. McMillan, The economic effects of fracture in the United
States, Tech. Report NBS special publication 647-2, United States Department
of Commerce, Washington, DC., 1983.
[36] D.S. Dugdale, Yielding in steel sheets containing slits, Journal of the Mechanics
and Physics of Solids 8 (1959), 100–104.
[37] F. Durbin, Numerical inversion of Laplace transofrms: an efficient improve-
ment to Dubner and Abate’s method, The Computer Journal 17 (1974), 371–
376.
BIBLIOGRAPHY 139
[38] W.M. Elleithy, H.J. Al-Gahtani, and M. El-Gebeily, Iterative coupling of BE
and FE methods in elastostatics, Engineering Analysis with Boundary Elements
25 (2001), 685–695.
[39] F. Erdogan, Fracture problems in composite materials, Engineering Fracture
Mechanics 4 (1972), 811–840.
[40] P. Fedelinski, M.H. Aliabadi, and D.P. Rooke, The dual boundary element
method in dynamic fracture mechanics, Engineering Analysis with Boundary
Elements 12 (1993), 203–210.
[41] , The dual boundary element method: J-integral for dynamic stress in-
tensity factors, International Journal of Fracture 65 (1994), 369–381.
[42] , A single-region time domain BEM for dynamic crack problems, Inter-
national Journal of Solids and Structures 32 (1995), 3555–3571.
[43] , Boundary element formulations for the dynamic analysis of cracked
structures, Engineering Analysis with Boundary Elements 17 (1996), 45–56.
[44] P. Fedelinski, M.H. Aliabadi, and D.P Rooke, The Laplace transform DBEM
for mixed-mode dynamic crack analysis, Computers and Structures 59 (1996),
1021–1031.
[45] P. Fedelinski, M.H. Aliabadi, and D.P. Rooke, The time-domain DBEM for
rapidly growing cracks, International Journal for Numerical Methods in Engi-
neering 40 (1997), 1555–1572.
[46] R. Gallego and J. Dominguez, Hypersingular BEM for transient elastodynamics,
International Journal for Numerical Methods in Engineering 39 (1996), 1681–
1705.
BIBLIOGRAPHY 140
[47] S. Ganguly, J.B. Layton, and C. Balakrishma, Symmetric coupling of multi-
zone curved galerkin boundary elements with finite elements in elasticity, Inter-
national Journal for Numerical Methods in Engineering 48 (2000), 633–654.
[48] P.H. Geubelle and M.S. Breitenfeld, Numerical analysis of dynamic debonding
under anti-plane shear loading, International Journal of Fracture 85 (1997),
265–282.
[49] A.A. Griffith, The phenomena of rupture and flow in solids, Philosophical
Transactions, Series A 221 (1920), 163–198.
[50] J. Hadamard, Lectures on Cauchy’s problem in linear differential equations,
Yale University Press, New Haven, 1923.
[51] G. Hasse, Parallel incomplete Cholesky preconditioners based on the non-
overlapping data distribution, Parallel Computing 24 (1998), 1685–1703.
[52] E. Heymsfield, Infinite domain correction for anti-plane shear waves in a two-
dimensional boundary element analysis, International Journal for Numerical
Methods in Engineering 40 (1997), 953–964.
[53] P.D. Hilton and G.C. Sih, A laminate composite with a crack normal to the
interfaces, International Journal of Solids and Structures 7 (1971), 913–930.
[54] C.H. Im, H.K. Kim, and H.K. Jung, Efficient technique for 3-D edge element
method considering geometrical symmetry, IEEE Transactions on Magnetics 37
(2001), 3190–3193.
[55] G.R. Irwin, Onset of fast crack propagation in high strength steel and aluminium
alloys, Proceedings of the Second Sagamore Ordinance Materials Conference
(1956), 289–305.
BIBLIOGRAPHY 141
[56] , Onset of fast crack propagation in high strength steel and aluminium
alloys, Sagamore Research Conference Proceedings 2 (1956), 289–350.
[57] , Analysis of stresses and strains near the end of a crack traversing a
plate, Journal of Applied Mechanics 24 (1957), 361–364.
[58] S.H. Kim, K.Y. Lee, and M.B. Park, Mode II stress intensity factor for lay-
ered material under arbitrary shear crack surface loading, Engineering Fracture
Mechanics 55 (1996), 85–94.
[59] E. Kita and N. Kamiya, Subregion boundary element method, JSME Interna-
tional Journal. Series A, Mechanics and Material Engineering 37 (1994), 366–
372.
[60] H.R. Kutt, On the numerical evaluation of finite-part integrals involving an
algebraic singularity, CSIRO Special Report WISK 179, National Research In-
stitute for Mathematical Sciences, 1975.
[61] L. Lapidus and G.F. Pinder, Numerical solution of partial differential equations
in science and engineering, John Wiley and Sons, New York, 1982.
[62] D.P. Linz, On the approximate computation of certain strongly singular inte-
grals, Computing 35 (1985), 345–353.
[63] N. Liu and N.J. Altiero, An integral equation method applied to mode III crack
problems, Engineering Fracture Mechanics 41 (1992), 587–596.
[64] X.P. Lu and W.L. Wu, A new subregion boundary element method, 15th Inter-
national Conference on Boundary Element Technology (2003), 115–124.
[65] , A subregion DRBEM formulation for dynamic analysis of two dimen-
sional cracks, Mathematical and Computer Modelling (Accepted).
BIBLIOGRAPHY 142
[66] , Dual boundary integral formulation for 2-d linear elastic crack prob-
lems, Journal of Computational Mathematics (Submitted).
[67] , A new subregion boundary element technique based on the domain de-
composition method, Engineering Analysis with Boundary Elements (Submit-
ted).
[68] S.W. Ma, A central crack in a rectangular sheet where its boundary is subjected
to an arbitrary anti-plane load, Engineering Fracture Mechanics 30 (1988), 435–
443.
[69] , A central crack of mode III in a rectangular sheet with fixed edges,
International Journal of Fracture 39 (1989), 323–329.
[70] S.W. Ma and L.X. Zhang, A new solution of an eccentric crack off the center
line of a rectangular sheet for mode-III, Engineering Fracture Mechanics 40
(1991), 1–7.
[71] J. Martinez and J. Dominguez, Short communication : On the use of quarter-
point boundary elements for stress intensity factor computations, International
Journal for Numerical Methods in Engineering 20 (1984), 1941–1950.
[72] H. Mews and G. Kuhn, An effective numerical stress intensity factor calculation
with no crack discretization, International Journal of Fracture 38 (1988), 61–76.
[73] Y. Murakami, Stress intensity factors handbook, Pergamon Press, Oxford, 1987.
[74] V. Murti, S. Valliappan, F. Asce, I.K. Lee, and M. Asce, Stress intensity factor
using quarter point element, Journal of Engineering Mechanics 111 (1985), 203–
217.
[75] N. Papamichel and G.T. Symm, Numerical techniques for two-dimensional
Laplacian problems, Computer Methods in Applied Mechanics and Engineering
6 (1975), 175–194.
BIBLIOGRAPHY 143
[76] P.W. Partridge, C.A. Brebbia, and L.C. Wrobel, The dual reciprocity boundary
element method, Computational mechanics publications, Southampton, 1992.
[77] G.H. Paulino, M.T.A. Saif, and S. Mukherjee, A finite elastic body with a curved
crack loaded in anti-plane shear, International Journal of Solids and Structures
30 (1993), 1015–1037.
[78] D.W. Pepper and J.C. Heinrich, The finite element method: Basic concepts and
applications, Hemisphere, United States of America, 1992.
[79] A. Portela, M.H. Aliabadi, and D.P. Rooke, Dual boundary element analysis
of cracked plates: singularity subtraction technique, International Journal of
Fracture 55 (1992), 17–28.
[80] , The dual boundary element method: effective implementation for crack
problems, International Journal for Numerical Methods in Engineering 33
(1992), 1269–1287.
[81] J.R. Rice, A path independent integral and the approximate analysis of strain
concentration by notches and cracks, Journal of Applied Mechanics (1968), 379–
386.
[82] J.R. Rice and G.C. Sih, Plane problems of cracks in dissimilar media, Journal
of Applied Mechanics (1965), 418–423.
[83] V. Rischmuller and W.M. Rucker S. Kurz, Parallelization of coupled differen-
tial and integral method using domain decomposition, IEEE Transactions on
Magnetics 38 (2002), 981–984.
[84] F.J. Rizzo, An integral equation approach to boundary value problems of classical
elastostatics, Quarterly of Applied Mathematics 25 (1967), 83–95.
BIBLIOGRAPHY 144
[85] A. Saez, R. Gallego, and J. Dominguez, Hypersingular quarter-point boundary
elements for crack problems, International Journal for Numerical Methods in
Engineering 38 (1995), 1681–1701.
[86] C. Shen and J. Zhang, Parallel two level block ILU preconditioning techniques
for solving large sparse linear systems, Parallel Computing 28 (2002), 1451–
1475.
[87] G.C. Sih, Stress distribution near internal crack tips for longitudinal shear prob-
lems, Journal of Applied Mechanics (1965), 51–58.
[88] J. Sladek and V. Sladek, Dynamic stress intensity factors studied by boundary
integro-differential equations, International Journal for Numerical Methods in
Engineering 23 (1986), 919–928.
[89] , A boundary integral equation method for dynamic crack problems, En-
gineering Fracture Mechanics 27 (1987), 269–277.
[90] M.D. Snyder and T.A. Cruse, Boundary-integral equation analysis of cracked
anisotropic plates, International Journal of Fracture 11 (1975), 315–328.
[91] P. Sollero, M.H. Aliabadi, and D.P. Rooke, Anisotropic analysis of cracks em-
anating from circular holes in composite laminates using the boundary element
method, Engineering Fracture Mechanics 49 (1994), 213–224.
[92] G.T. Symm, Treatment of singularities in the solution of laplace’s equation by
an integral equation method, Tech. Report NAC31, National physical laboratory,
1973.
[93] H. Tada, P. Paris, and G. Irwin, The stress analysis of cracks handbook, Paris
Productions Incorporated, 1985.
BIBLIOGRAPHY 145
[94] K. Ting, K.K. Chang, and M.F. Yang, Analysis of mode III crack problems
by boundary element alternating method, Fatigue, Flaw Evaluation and Leak-
Before-Break Assessments, ASME 280 (1994), 21–26.
[95] K. Weeber and S.R.H. Hoole, The subregion method in magnetic field analysis
and design optimization, IEEE Transactions on Magnetics 28 (1992), 1561–
1564.
[96] A.A. Wells, The condition of fast fracture in aluminium alloys with particular
reference ot comet failures, British Welding Research Association Report NRB
129 (1955).
[97] , Application of fracture mechanics at and beyond general yielding,
British Welding Journal 10 (1963), 563–570.
[98] P.H. Wen, M.H. Aliabadi, and D.P. Rooke, Cracks in three dimensions: A
dynamic dual boundary element analysis, Computer Methods in Applied Me-
chanics and Engineering 167 (1998), 139–151.
[99] P.H. Wen, M.H. Aliabadi, and A. Young, Dual boundary element methods for
three-dimensional dynamic crack problems, Journal of Strain Analysis for En-
gineering Design 34 (1999), 373–394.
[100] , Stiffened cracked plates analysis by dual boundary element method, In-
ternational Journal of Fracture 106 (2000), 245–258.
[101] H.M. Westergaard, Bearing pressures and cracks, Journal of Applied Mechanics
6 (1939), 49–53.
[102] M.L. Williams, On the stress distribution at the base of a stationary crack,
Journal of Applied Mechanics 24 (1957), 109–114.
BIBLIOGRAPHY 146
[103] L.S. Xanthis, M.J.M. Bernal, and C. Atkinson, The treatment of singularities
in the calculation of stress intensity factors using the integral equation method,
Computer Methods in Applied Mechanics and Engineering 26 (1981), 285–304.
[104] C. Zhang and D. Gross, A non-hypersingular time-domain BIEM for 3-D tran-
sient elastodynamic crack analysis, International Journal for Numerical Meth-
ods in Engineering 36 (1993), 2997–3017.
[105] X.S. Zhang, The general solution of a central crack off the center line of a
rectangular sheet for mode III, Engineering Fracture Mechanics 28 (1987), 147–
155.
[106] , A tearing mode crack located anywhere in a finite rectangular sheet,
Engineering Fracture Mechanics 33 (1989), 509–516.
[107] Z.G. Zhou and L. Ma, Two collinear Griffith cracks subjected to anti-plane
shear in infinitely long strip, Mechanics Research Communications 26 (1999),
437–444.
[108] O.C. Zienkiewicz and R.L. Taylor, The finite element method. (Fourth edition),
McGraw-Hill, New York, 1989.